The Net Zero Game

Share of 2020 private-car demand still served by a liquid-fuelled car in 2050. The four car shares are rebalanced to 100% as the player moves them.

Car demand transferred to short-distance rail, at the occupancy and unit consumption of the rail row rather than the car row.

Flagged above 25%. No published French trajectory, négaWatt's included, cuts passenger travel by much more than a quarter. Past 25% the scenario rests on a change in where people live, work and go that nothing in this model brings about, and whose cost — in housing, in services, in time — appears nowhere in it.

Flagged above 25%. A quarter less freight is a different economy, not a more efficient one: tonne-kilometres follow what a country makes, imports and consumes. Past 25% the scenario assumes that change rather than producing it — and what the country stops making is not re-imported anywhere in this account.

In 2050 the model leaves no fossil liquid fuel at all: every litre is either biofuel or e-fuel made from electricity. This is a scenario assumption, and it is why liquid fuel carries a low emission factor.

The headline decarbonisation choice for buildings, and the one that drives the winter peak. It is a share of heat need, not of energy: how that heat is produced is the next question down.

A target in the unit the resource constraint is written in, so it can be read straight against the biomass limit instead of being reconstructed from two shares. 46 TWh of wood delivers 39 TWh of heat at a boiler efficiency of 0.85, which is roughly what the previous scheme produced.

Seasonal COP 2.5, falling to 2.0 at peak.

Seasonal COP 3.0, falling to 2.0 at peak.

A slider rather than a stock that can only shrink, because a scenario may genuinely install more electric convectors — they are cheap to fit and terrible for the peak. Efficiency 1 in both seasons, so this is the one electric option that buys no COP at all.

Declared in TWh so that it, the network heat pumps and the recovered heat together say how much of the network is decarbonised, and gas absorbs whatever is left.

Industrial waste heat, incineration and geothermal. It has no emission factor and adds nothing to the winter peak, which makes it the cheapest thing a network can run on — and the model does not check it against the waste-heat gisement the industry module computes, so raising it far is optimistic in a way nothing here will stop you from being.

One slider conflates retrofit depth and retrofit rate, which have very different costs. Separating them is a documented next step.

Flagged above 50%. Halving the average heating demand of the whole stock means bringing essentially every dwelling to a level France currently reaches a few tens of thousands of times a year, every year until the horizon. One slider also conflates retrofit depth and retrofit rate, whose costs are very different, so past this point it hides which of the two is being asked for.

Residential floor area started, from the ministry's own construction statistics: 20.4 Mm² in 2024 and 21.0 in 2025, against 34.7 in 2021. France started 412 600 dwellings in 2021 and 258 100 in 2024, a fall of 37% in three years, and the average new dwelling has shrunk from 83 to 76 m² over the same period. The default is the 2024 figure because it is the most recent consolidated year, and because a model whose reference sits on a 2021 peak would make every sufficiency scenario look easy. The range is an argument, not a measurement. The low end, 8 Mm²/y, is roughly where the official housing-need study lands for the 2040s: the statistical service's central scenario needs 208 000 additional main residences a year in the 2020s, 139 000 in the 2030s and 55 000 in the 2040s, because household growth falls from +215 000 a year to +27 000 by 2045-2050 as household size drops to 1.99. The high end, 36, is above anything France has built this century. The draft national strategy assumes 310 000 dwellings a year to 2030 and 100 000 a year over 2040-2050, which straddles the middle of this slider. What it does not include is the sufficiency answer that needs no construction at all: France holds 3.0 million vacant dwellings, 1.2 million of them vacant for over a year, and 3.7 million second homes. The statistical service reckons a 3% vacancy floor in every employment zone would release 600 000 of them. This model has no lever for that, and the omission is named.

• SDES, Sit@del2 — logements commencés et surface de plancher, séries 2021-2025 (20.4 Mm² en 2024, 21.0 en 2025)
• SDES, «Besoins en logements à l'horizon 2050» (juin 2025) — 208 000 / 139 000 / 55 000 résidences principales par an
• Insee, projections de ménages 2018-2050 — +215 000/an jusqu'en 2030, +84 000/an ensuite, taille des ménages 2,20 → 1,99

Flagged above 30 Mm²/y. Above 30 Mm²/y the scenario builds through the 2030s and 2040s at the rate of the 2021 peak, into a country whose household growth has fallen by a factor of eight. The official need study reaches 55 000 additional main residences a year by the 2040s; this is about six times that.

Non-residential floor area started, all destinations: 19.9 Mm² in 2025, against about 31 in 2019. The destination split is the reason this lever is not called "tertiary": public buildings 4.01, warehouses 3.77, retail 3.54, farm buildings 3.44, offices 2.61, industry 2.51. Two fifths of it is warehouses and agriculture — floor area that carries cement, carries structural steel, and carries almost no heating. Tertiary proper, the part the building stock counts, is about 9.6 Mm²/y of the total. That distinction is why stage A stops where it does. The cement follows the whole 19.9; the heat would follow only the tertiary half, and the building module has no construction flow to put it in. See the module's own why.

• SDES, Sit@del2 — locaux non résidentiels commencés par destination, 2025 (19.9 Mm², dont entrepôts 3.77 et exploitations agricoles 3.44)

The default is the observed share, derived in timber_share_base and equal to it by construction — a test asserts the two agree, because a scenario that starts anywhere else would book a cement saving for buildings France has already put up. The range spans the whole published argument, and the two ends are not equally supported. ADEME's biosourced-development scenario, the more cautious of the two French projections, cuts cement by 2% over 2015-2034 and 8% over 2035-2049 — a much weaker result than the phrase "build in timber" suggests. négaWatt goes to 80% of detached houses by 2030 and 95% by 2050, with 40% on collective housing and tertiary, and reports −46% of concrete and −60% of clinker. Both are in the literature; they differ by a factor of five, and the maximum here is négaWatt's order of magnitude rather than ADEME's. Two limits the slider cannot express and the annex has to. A timber building is not a building without concrete: the measured case this module's coefficient comes from still carried 288 kg/m² of it in foundations and ground floor. And the wood has to come from somewhere — at the top of this slider new-building structure alone asks for 6.1 Mm³ of sawn product a year, against a French softwood sawnwood production of about 7.0 Mm³, and land_timber_headroom shows what that leaves the rest of the long-lived harvest.

• Enquête nationale de la construction bois, activité 2024 — 6.6% du logement neuf, 17.6% de la surface non résidentielle neuve
• ADEME 2019, «Prospective de consommation de matériaux pour la construction des bâtiments» — scénario biosourcé : ciment −2% (2015-34) puis −8% (2035-49)
• négaWatt 2022, rapport complet partie 4 — 95% des maisons et 40% du collectif et du tertiaire en bois en 2050 ; béton −46%, clinker −60%
• IGN–FCBA, «La forêt en 2050» (2024) — l'offre supplémentaire de grumes résineuses reste inférieure de 1.0 à 1.5 Mm³/an à la demande en 2050

Flagged above 45%. At 45% the structure of new buildings alone asks for 3.4 Mm³ of sawn product a year — half of everything French sawmills cut from softwood — in a country that already imports a quarter of the softwood sawnwood it uses and whose forest inventory projects additional sawlog supply falling short of additional demand by one to one and a half million cubic metres a year in 2050, even in its increased-harvest cases. Nothing here stops the slider: what it cannot supply it imports, and imported timber stores no carbon in the French inventory, because the harvested-wood-products pool runs on the production approach. Past this point the cement saving is real and the sink belongs to somebody else.

A rule, and the honest kind: the model has no driver for a road programme, so the volume of civil works is set by hand and the annex says so. 100 is "as much as today". What it moves is a third of French cement — roads 13%, buried networks 13%, bridges and retaining structures 9% on the sector's own 2018 end-use map. That is the answer to the question this module was built to answer: a floor-area lever cannot reach it. Concrete's share in civil works is also the one place the timber literature holds constant, because its properties are hard to replace there, so this third is out of reach of the timber lever too. The bounds are a judgement. Civil-works output grew 4% in 2024 while building collapsed, and the network-renewal backlog argues for more rather than less; but nothing published sets a 2050 volume, and a slider that could halve French road building without a word would be worse than a stated rule.

• SFIC / CIMbéton / ATILH, carte des usages du ciment 2018, reprise par The Shift Project, «Décarboner la filière ciment-béton» (janvier 2022), figure 6 — routes 13%, VRD 13%, ouvrages d'art 9%
• France Stratégie, «Les coûts d'abattement — Partie 6 : Ciment» (mai 2023) — «utilisé à presque deux tiers par le bâtiment et à un peu plus d'un tiers par les travaux publics»
• FNTP, «L'activité du secteur en chiffres» — travaux publics 51.3 Md€ en 2024, +4%

Not shown in this edition. This country's package hides the control; the model reads the lever at the default in the next column.

Retired in France since the food module arrived, and hidden. The ammonia tonnage is now mineral_nitrogen × ammoniaDomesticShare ÷ 0.822 plus the ammonia French chemistry makes for something other than fertiliser, so a fertiliser decision is a hydrogen decision and the industry chain reads a demand instead of an assertion. At the base year's 1 817 kt of nitrogen and today's 34% domestic share the derivation gives 900.2 kt — this slider's own value, reproduced to a fifth of a kilotonne from two numbers that knew nothing about it. At the reference nitrogen dose it gives 509 kt, and the hydrogen the chain draws falls with it. It stays declared, at the value it always had, because the three provisional editions still use it and because a hidden lever that still moved a number would be a decision the player cannot see; a test asserts it is inert here. The entry itself was not restructured — moving it out of the shared model would have been a change to the country contract, and the same result is had with the land_module_active switch that already gates the rest of the module. What it used to be: one figure instead of a green/grey split. Until v0.12.0 ammonia was two rows — 700 kt made from electrolytic hydrogen and 200 kt from a reformer — which put the hydrogen route inside the ammonia lever and nowhere else. Now every tonne consumes the same 5.94 MWh of hydrogen and the hydrogen mix decides how it was made, which is where that decision belongs: the same reformer serves steel, refining and everything else.

• Teaching workbook, "Industry" sheet — 700 kt green plus 200 kt grey

Only the biogenic fraction of the CO₂ fed to the synthetic-olefin route counts as a removal, which is why this lever alone can turn the olefin process term negative.

How much less plastic the country asks for, against the base year — a demand reduction and not a recycling rate. It scales the olefin tonnage the crackers make, and it is the one lever in this chain that answers "how much of this do we need" rather than "how do we make it". The two French reference scenarios cannot anchor it. ADEME's Transition(s) 2050 publishes material-demand trajectories for steel, aluminium, cement and glass and none for plastics; its plastics content is an 80% recycling rate, and négaWatt does the same. Both express plastics as a rate of recycling, and this model has no recycling variable, so neither can be read onto this slider. What can is SYSTEMIQ's ReShaping Plastics, the one published pathway that separates the wedges. European plastic demand grows in its baseline, 37 to 48 Mt by 2050; its Circularity scenario takes 25% off that baseline by reducing demand and 4% more by substituting materials. Against 2050 that is 29%; against today it is under 10%, because a quarter of a growing baseline is a tenth of the present. This slider is written against the base year, so the two readings are the two ends of what one study supports — and the reference, 30%, sits at the far edge of the more generous one. The sectors move in opposite directions underneath: packaging can lose 38% (an eighth eliminated, a further three tenths reused), vehicles 22%, while construction plastic grows by half in every scenario published. A single national share hides that, and a player moving this slider is assuming the packaging wedge does all the work.

• SYSTEMIQ (2022), ReShaping Plastics: Pathways to a Circular, Climate Neutral Plastics System in Europe, for Plastics Europe — EU27+UK, baseline 37→48 Mt by 2050; Circularity scenario 25% reduce + 4% substitute
• ADEME, Transition(s) 2050 — material-demand trajectories for steel, aluminium, cement and glass; plastics carried as a recycling rate
• See docs/waste/olefins_and_plastic_demand.md for the legal instruments (PPWR, SUP, loi AGEC, décret 3R) and why they do not bound this slider

Flagged above 30%. Past 30% no published pathway supports this as a demand reduction. SYSTEMIQ's most circular European scenario takes 29% off a 2050 baseline that has grown by a third — under a tenth against today — and 30% here is already the far edge of reading that against the base year instead. The 70% at the top of this slider exists in the literature only as three quarters less virgin fossil plastic, and that figure already contains the recycling, the material substitution and the CO₂ and bio feedstock which this model carries separately in olefinRoute and biogenicCO2 — so reading it here counts the same effort twice.

Renamed in v0.20, because it finally has a driver. Until then it was "cement-demand reduction" with no why, no help and nothing behind it: a scenario could take 60% off the country's cement without saying which building it had not built, and the annex could not say what the slider meant. How much is built is now the construction module's business. What is left here is the intensity of the works: leaner mixes, thinner slabs and post-tensioned structures, a design that uses the concrete it pours. It is not clinkerRate, which replaces clinker inside the cement with slag or fly ash; this one asks for less cement in the first place. The two multiply, and a scenario that pushes both is asking for a lot from the same building twice over. The range is a rule. The published levers for the sector are the clinker ratio and capture, and nobody has assessed how far structural design alone can go — so 60% is a bound on an argument rather than on an assessment.

The default is the industry's own 2050 target, near enough: the sector roadmap takes the average clinker content of French cement from 77% in 2015 to 68% in 2030 and 62% in 2050. The two are not quite the same quantity. The roadmap's content counts imported clinker, and this rate turns cement demand into French clinker — 71.6% in 2020, where the content was about 78, the difference being clinker fired abroad. 60 sits between the published target and what it would be net of imports. 78, the maximum, is today's content; 35 is a cement that is mostly slag, calcined clay and limestone, and the flag below 45 says why that needs materials France may not have.

• Conseil national de l'industrie / DGE (May 2023), Feuille de route de décarbonation de la filière Ciment, p. 13 and Tableau 3-4 — clinker content 77% (2015), 68% (2030), 62% (2050)
• France Ciment / SFIC, Le ciment en chiffres 2022, p. 4 — clinker 11 759 kt, cement 16 422 kt in 2020

Flagged below 45%. A lower clinker rate needs something to put in the clinker's place, and the two supplementary materials that work at scale are blast-furnace slag and coal fly ash — both disappearing from Europe at exactly the moment this scenario asks for more of them. Below 45% the fleet-average cement depends on an addition nobody has shown France will have.

The share of the kiln's calcination CO₂ that is captured. A real capture plant on a cement stack takes the fuel CO₂ with it, so this lever is a floor on what capture delivers, not an estimate of it: in a French works today the kiln fuel is about a third of the stack. Until 0.27.0 the term it multiplied carried the kiln fuel too, which made the lever look like a whole-stack capture while the same fuel was also being charged through the energy account — emitted twice, captured once.

Flagged above 60%. No French cement works captures its CO₂ today, and the one project to have reached a final investment decision covers well under a tenth of the national clinker line. Past 60% the scenario has fitted, powered and paid for capture across most of the fleet inside twenty-five years — and the model charges it for none of the three.

How much the branches the game does not model produce, between today's output and the source workbook's 2050 scenario. That scenario is a reindustrialisation: measured branch by branch it multiplies textile output by 8.5, electronics by 3.1, mineral extraction by 2.5, while mineral chemistry falls to 0.52 and naval and aerospace to 0.56. Growth of that size is not decarbonisation, and separating it from the process lever is the whole point — otherwise a player could appear to clean up industry while assuming an eightfold textile sector.

• Offre_et_demande.xlsx, branch sheets E18–E38 — production 2019 and 2050 by product and process
• INSEE, enquête EACEI 2019, for the observed energy consumption by branch

How much energy each unit of output takes, between today's processes and the 2050 ones. This is where electrification lives: at constant output it takes the electricity of these branches from 68 to 138 TWh while cutting coal and fuel oil to zero. The default is 100% and the volume default is 0%, so the reference scenario reads "the rest of industry modernises at today's output" — the neutral reading, with growth added explicitly.

• Offre_et_demande.xlsx, branch sheets E18–E38 — unit consumption in MWh per tonne, 2019 and 2050

Heat rejected by industrial processes and recovered instead of vented. ADEME puts the recoverable gisement at 15.6 TWh on the 2019 industry, of which 7.7 TWh above 100 °C, and — this is what makes it a real constraint — expresses it against the fuel each sector burns. So the gisement is not a fixed reserve: it shrinks as processes electrify, because there is no combustion left to reject heat from. Recovering waste heat and electrifying heat compete for the same physics, and the model makes them compete.

• ADEME, La chaleur fatale (2017 study, 2021 update)

Cross-cutting energy efficiency — motors, drives, compressed air, insulation, heat integration. RTE, after CEREN, identifies a potential of 21.1% on industrial electricity and 19.8% on fuels, of which about 58% pays back in under three years. The ceiling is branch-specific and applied as such: 11.4% for steel, 15.5% for metals and machinery, 19.3% for paper, 24.1% for minerals, 25.0% for the food industry, 31.1% for chemistry. This lever says how much of that identified potential is actually captured, not how much exists — so it cannot invent efficiency beyond what the study found, which is the point of a ceiling.

• RTE, Bilan prévisionnel — groupe de travail Consommation, after CEREN

Flagged above 75% of the identified potential. About 58% of the potential RTE identifies after CEREN pays back in under three years. Past 75% the scenario is taking efficiency nobody has shown pays for itself, in a model that prices none of it. The slider says how much of the identified potential is actually captured, so taking almost all of it is an assumption about industrial investment rather than about engineering.

The model assumes all 2050 methane is biomethane, so it carries a life-cycle factor rather than the 227 gCO₂/kWh of fossil natural gas. Raising this slider towards 227 shows what happens if the biomethane assumption fails. 25 sits inside the published range, and the range is wide. The two French life-cycle studies of the biomethane mix give 23.4 gCO₂/kWh (multifunctional allocation) and 44 (attributional); the renewable-energy directive's own defaults run from −302 for biomethane from wet manure with a closed digestate store, where the avoided storage methane is credited, to +184 for maize whole-plant. Leakage is 1–6% of the methane nominally and has been measured far higher. So the factor is a statement about the feedstock mix as much as about the process — a manure-heavy supply would justify a lower number and a crop-heavy one a higher — and since stage C the model knows the mix well enough for a reader to make that argument, even though the factor does not respond to it.

• Teaching workbook, "Synthesis" sheet S15 and S42
• Sénat, Méthanisation : au-delà des controverses, r20-872 (2021) — Quantis/GRDF 23,4 et 44 gCO₂e/kWh, fuites 1–6 %
• Directive (UE) 2018/2001, annexe VI — −84 g/MJ (lisier, digestat couvert), +51 g/MJ (maïs plante entière)

Same logic as methane: the 2050 model leaves no fossil liquid fuel, so the factor is that of biofuel and e-fuel, against 264 gCO₂/kWh for the 2020 fossil fuel it replaces. It kept its value when the biofuel supply became a computed one, and the reason is where the cultivation goes. The renewable-energy directive's default for rapeseed FAME is 179 gCO₂e/kWh, of which about 115 is cultivation — fertiliser N₂O and the upstream nitrogen. This model books that nitrogen in agriculture, as the national inventory does: the hectares energyCropArea names are cropland inside the same arable area the crops block fertilises, so charging the fuel for them as well would count one tonne of N₂O twice. What the fuel carries is the processing and transport, which is roughly 60–65 gCO₂/kWh for rapeseed FAME and 54 for waste-oil HVO — still above 25, and the gap is the model's assumption that most of the 2050 liquid is e-fuel made on clean electricity rather than crop oil. Two limitations, stated rather than hidden. The model does not split the factor by origin, so a scenario that meets its liquid demand entirely on 1G crops is charged the same as one that meets it on e-fuel. And expanding energyCropArea adds no mineral nitrogen: the dose is an intensity on an arable area held at the base year's, so a fuel crop displaces a food crop on land already fertilised. The crop block counts that displaced food — the fuel hectare is arable land the plates, the herd and the exports can no longer use — but it does not fertilise the fuel hectare itself, and that is the half of the question still open.

• Teaching workbook, "Synthesis" sheet S17 and S44
• Directive (UE) 2018/2001, annexe V — EMAG de colza 179 gCO₂e/kWh par défaut (162 typique), culture seule 32,0 g/MJ ; HVO d'huile usagée 54
• docs/agriculture/bioenergy.md §4.4, option (a) — la culture reste dans l'agriculture, le carburant porte la transformation

27 gCO₂/kWh, the same figure the workbook observes for 2020, rather than the zero its 2050 column carries. The zero is the biogenic-carbon convention: burning wood emits CO₂, but the convention books it against the forest that regrew rather than against the boiler. Applying it to one year and not the other made the two ends of the model incomparable — wood appeared to decarbonise between 2020 and 2050 without anything physical changing. Holding the factor constant means a scenario that leans on wood is charged for it in both years, and moving this slider to zero still shows exactly what the convention is worth: 2.06 MtCO₂ at the reference scenario, 1.6 of it in buildings. Since stage C the convention is no longer free. The combustion CO₂ the factor omits reappears in the forest sink's response to harvest: cutting more wood feeds the boiler and costs the sink, in the same scenario and from the same identity in cubic metres. A wood-heavy scenario can no longer borrow from a forest the model was not looking at.

• Teaching workbook, "Synthesis" sheet S16 and S43
• The 27 gCO₂/kWh is the workbook's own 2020 value; it covers the fossil energy of the wood chain, not the combustion CO₂ the convention omits.

Kerosene per passenger-kilometre, improving each year through aircraft renewal, seat density and load factor. Published trajectories converge tightly on 1%/year: ICAO 1.0, ADEME 1.0, T&E 0.9, the UK Committee on Climate Change 0.9, the World Economic Forum's Clean Skies for Tomorrow 1.0, against 2.5 in the more optimistic ICSA figure. The default is 0 because the source workbook uses today's consumption for 2050 — moving the slider to 1 shows what a quarter-century of fleet renewal is worth, and it is worth less than most people expect.

• ICAO, Environmental trends in aviation to 2050 (2019)
• ADEME, Élaboration de scénarios de transition écologique du secteur aérien
• Transport & Environment, Roadmap to decarbonising European aviation (2018)

Flagged above 1.5%/y. Published trajectories converge tightly on 1%/year — ICAO 1.0, ADEME 1.0, T&E 0.9, the UK Committee on Climate Change 0.9, the World Economic Forum's Clean Skies for Tomorrow 1.0 — against 2.5 in the most optimistic ICSA figure. Past 1.5%/year the scenario assumes a quarter-century of fleet renewal faster than any of them, compounded over the whole horizon.

Passenger-kilometres, compounded over thirty years. The source workbook carries 2020 demand straight through to 2050, which is a growth assumption of zero and a very strong one — no published trajectory says that. For flights departing France the DGAC roadmap gives 1.62%/year falling to 1.19, or 1.8 without a price effect and 0.8 with one; ADEME spans −1.3 to +3.0 depending on scenario and price effect; Eurocontrol gives 3.3 then 1.7. World figures are higher still: ICAO 1.1 to 3.4, Airbus 2.6, Boeing 5.6 falling to 2.5. At 1.5%/year over thirty years traffic grows by 56%, which is worth more than every efficiency gain in the sector combined.

• DGAC, feuille de route de décarbonation de l'aviation
• ADEME, Élaboration de scénarios de transition écologique du secteur aérien
• ICAO environmental trends to 2050; EASA; Eurocontrol; Airbus and Boeing market outlooks

Flagged above 2.5%/y. 2.5%/year compounds to 2.1 times today's traffic by 2050. For flights departing France the DGAC roadmap gives 1.62%/year falling to 1.19, and ADEME spans −1.3 to +3.0 across its whole scenario set. Past 2.5 the scenario is on the world's most bullish manufacturer forecasts, applied to a country whose own regulator does not use them.

Sustainable aviation fuel from biomass. The published estimates disagree by a factor of six, and which route is assumed matters as much as who estimated it: HEFA from waste oils 600–1 900 €/t, biomass-to-liquid 1 400–2 900, alcohol-to-jet 750–3 900. The default sits mid-range across the three. The width of that range is the honest answer, which is why this is a slider and the range is printed beside it.

• EASA, State of the EU SAF market; European Commission ReFuelEU impact assessment
• ISAE-Supaero, Cost estimation of the use of low-carbon fuels in prospective aviation
• ATAG Waypoint 2050; IEA Advanced Biofuels; Solakivi et al. (2022); O. Massol et al. (2025); DGAC decarbonisation roadmap

Synthetic kerosene from electrolytic hydrogen and captured CO₂. Estimates range from 1 820 €/t (European Commission) to 10 000 (DGAC), with the review's central band at 3 700–6 200. Direct air capture costs more than biogenic CO₂: EASA gives 7 300–8 700 €/t for atmospheric CO₂ against 6 600–7 975 for biogenic.

• EASA and European Commission ReFuelEU Aviation impact assessment
• ISAE-Supaero; ATAG Waypoint 2050; E-Cube; Brynolf et al. (2020); Solakivi et al. (2022); O. Massol et al. (2025)

Everything that is not fuel — aircraft, crew, airport charges, maintenance, overheads — is assumed unchanged in 2050 and is derived from today's ticket through this share. It is the weakest link in the ticket calculation: a 2050 airline may well have a different cost structure, and nothing here models that.

• EASA, European Aviation Environmental Report 2025
• European Commission, study supporting the ReFuelEU Aviation impact assessment (19–33%)

Not shown in this edition. This country's package hides the control; the model reads the lever at the default in the next column.

Retired in France since the food module arrived, and hidden. The agriculture sector is now built the way transport, building, industry and energy already are — as a sum of the post table, from three constructive rows: a herd sized by what the country eats and exports, a nitrogen balance on the fields, and the fuel the farm burns. This slider moves nothing here, and its help says so rather than describing a trajectory nobody is on. It stays declared, at the value it always had, because land_module_active reads it and because the three provisional editions still use it: they carry no food module, so their agriculture is still a position on a published trajectory between the observed year and the strategy's horizon. A test asserts that it is inert in France, which is a stronger statement than "it is not drawn". What it did, and why it had to go: it slid the sector between the 77.53 MtCO₂e observed in 2024 and the SNBC 3's 43.67 in 2050 with no driver at all. A player could not ask what a smaller herd, a different diet or half the nitrogen would do, because none of those were in the model — the slider was the model. Its 100% default also meant the reference scenario asserted the strategy's own result instead of producing one, and the module that replaced it lands about three megatonnes above that result, which is the information the slider was hiding.

Not shown in this edition. This country's package hides the control; the model reads the lever at the default in the next column.

Retired in France since the land module arrived, and hidden. The natural sink is now computed — seven land classes, six inventory pools and a forest identity in cubic metres — so this slider moves nothing here. It stays declared, at the value it always had, because land_module_active reads it and because the three provisional editions still use it as their whole land-use model; a test asserts that it is inert. Everything below is what it said while it was the model, and it is kept because it is the argument the module had to answer. Set directly rather than as a share of a trajectory, because the number is what a reader argues about. Declared as a magnitude absorbed, and negated in the equations: a slider whose right-hand end weakened the sink read backwards, since dragging right is dragging towards more. The dashboard still shows it signed, as −23. 23 is where the SNBC 3 pathway lands, and it is a weakening: the French forest sink has roughly halved since 2010 as the stock ages and dieback and drought bite. Reaching the stronger end of this range means a forest growing faster than it does today, and the model does not say what it would take to get there — that is a genuine gap, and one where defensible figures are scarce.

• SNBC 3 — puits de carbone des terres, trajectoire 2050

Set directly, for the same reason, and it is the single largest assumption in the whole model: megatonnes a year of capture and storage that nothing here builds, powers or pays for. Its cost — in euros and in the energy the capture itself consumes — is not modelled, so moving this slider is free in a way it would not be in reality. The default is no longer the closure residual, and that is deliberate. Until v0.19 it was 43: the difference between the −66 MtCO₂e of total 2050 absorptions the SNBC 3 publishes and the −23 of natural sink stated beside them, which is arithmetic rather than an assessment of anything. It is now 30, roughly France's share of the 280 MtCO₂e a year of injection capacity the European industrial carbon management strategy projects for 2050 across the Union — still a quantity nobody has built, but one argued from a published deployment rather than from a subtraction. The thirteen megatonnes that leaves are not hidden. The reference scenario no longer reaches the published total; the national reconciliation shows the gap against official_technological_sink_2050, which stays at the −43 the strategy implies; and the flag below is lit at the reference rather than only at the extremes. A player who wants the strategy's own closure drags the slider to 43 and reads what it stands on.

• Inferred 2050 closure value; see the national-reconciliation annex
• European Commission, Industrial Carbon Management Strategy (COM/2024/62)

Flagged above 20 MtCO₂e/y absorbed. Twenty megatonnes a year, for France alone, is around half of everything the planet currently captures and stores, across every facility in operation. France captures none of it today. Nothing in this model builds the plant, supplies the electricity the capture consumes, or pays for either — so every megatonne dragged past this point is free here and is free nowhere else.

Measured on Teruti, not on the cadastre, and the two differ by a factor of two or three: Teruti counts every garden and verge as artificialised and the cadastre counts parcels newly built on, which is why the country is at 15–20 kha/y on one measure and 38–52 on the other. The account is written in Teruti hectares, so it has to be moved by Teruti flows — and the emission content of the artificial pool only closes on Teruti's rate: 5.0 MtCO₂e ÷ 52 kha/y = 96 tCO₂ per ha/y of flow, inside the 15–110 tCO₂/ha range the underlying soil and biomass arithmetic gives. The research note recommended the cadastral measure as "the honest compromise" for a lever; we depart from it, for that reason, and the cadastral measure and the ZAN test belong beside the result as a comparison rather than inside it as the driver. The default is a halving of the 2011–2021 decade, which is what the Climat et Résilience law asks for by 2031 — roughly 12 kha/y on the cadastre, and carried across to the Teruti rate here, which is an approximation and is the weakest step in this lever. Zero is net-zero artificialisation reached early.

• Agreste, Sols artificialisés : +66 % en 40 ans, Primeur 2025-1 — 52 kha/y mean 1982–2022, under 38 recently
• Cerema, Analyse de la consommation d'ENAF 2011–2024 — 22.9 kha/y mean on the cadastre, 15.1 in 2024
• Loi n° 2021-1104 (Climat et Résilience), article 191 — half the previous decade by 2031, no net artificialisation in 2050

The two ends of this slider are not the same measurement, and that is the single largest unreconciled flow in the whole account. The SNBC 3 plans deliberate afforestation outside existing forest: 100 ha/y in 2021, rising to 15 000 by 2030, then 200 000 ha over 2030–2039. IGN's inventory measures the forest expanding by 90 000 ha/y, most of it spontaneous — canopy closing over abandoned heath and grazing. Teruti, on the same territory, sees about 35 000. IGN counts a closing canopy as forest while Teruti still sees heath, and no published concordance settles it. Hectares planted are booked at the expansion storage rate, 3.0 tCO₂/ha/y, and only once they are more than ten years old — new forest does not store like mature forest, and at 15 kha/y the whole term is worth 0.7 MtCO₂/y against a forest pool of about 20. Planting is slow, and this lever says so.

• IGN, Mémento de l'inventaire forestier 2024, p. 18 — forest expanding by 90 kha/y
• Projet de SNBC 3, 31 October 2024, p. 92 — afforestation outside forest, 100 ha/y (2021) to 15 000 (2030)
• Agreste, Primeur 2025-1 — Teruti sees the forest growing by about 35 kha/y over 1982–2022

Flagged above 35 kha/y. The SNBC 3 plans deliberate afforestation at 15 000 ha a year by 2030 and 200 000 ha over the decade after it — about 20 000 a year. Past 35 000 the scenario is no longer planting: it is claiming the 90 000 ha a year IGN measures the forest gaining on its own, as the canopy closes over abandoned heath and grazing. That expansion is nobody's policy, no budget line reaches it, and Teruti sees a third of it.

France ploughed 2.3 Mha of permanent grassland between 1982 and 2018, about 64 kha/y, and the SNBC 3 assumes the area is held from here on. Zero is therefore the strategy's position and not an observation of a stable countryside; the negative end is re-grassing, which no French scenario plans at that rate and which the module allows because a diet scenario in stage B will free the hectares for it. The soil-carbon consequence is asymmetric and that is the teaching in this lever: ploughing grassland loses about 1.0 tC/ha/y for twenty years, putting arable land back to grass gains about 0.5 — loss is twice as fast as gain. At +100 kha/y the twenty-year tail alone is 7.3 MtCO₂/y, which is more than half of what every soil-carbon practice in the INRAE study could store. The fastest way to lose the soil carbon debate is to plough.

• Agreste, L'occupation du sol entre 1982 et 2018, Dossiers 2021-3 — permanent grassland −2.3 Mha
• SNBC 3 — permanent grassland maintained to 2050 (7 150 kha 'productive' in the June 2026 text, 9.6 Mha in the November 2024 draft)

100% is 4.72 MtC/y — cover crops 2.02, in-field agroforestry 1.10, temporary grassland in rotations 0.76, grassland intensification 0.69, hedges 0.15 — which is 17.3 MtCO₂/y. No-till is deliberately excluded, because INRAE's own reading is that it redistributes carbon down the profile rather than adding any; putting it back would add 2.5 MtCO₂/y and the widely quoted "+21 MtCO₂/y, 4 per 1000" headline includes both it and forest land. The split between the two land uses follows the itemised practices — 14.777 on arable, 2.530 on grassland — rather than the areas they sit on, which is what the module specification proposed; the specification's "arable 16.0 / grassland 3.8" sums to 19.8, the potential with no-till, not to the 17.3 it also states. The total is the same either way, so only which pool the chart shows it in changes. 30% is a judgement, not a published trajectory: the SNBC 3 asks agricultural soils to "approach equilibrium by 2050" without saying what share of the potential that is. Everything here is a thirty-year rate on a soil that saturates, and after that it stops.

• INRAE, Stocker du carbone dans les sols français ? (2019), résumé, table 2

Flagged above 70%. 100% is INRAE's whole itemised potential — cover crops, in-field agroforestry, temporary grassland in rotations, hedges — taken on every eligible hectare at once. Past 70% the scenario has seven eligible hectares in ten actually converted and held for the whole horizon, which no French incentive scheme has ever approached, and the account charges nothing for the ones that lapse.

53.1 Mm³/y is what IGN measures as removals of live trees over 2014–2022, and it is not the 38 Mm³ of the commercial harvest statistic: the difference is the firewood that is cut and burned without ever being sold, about 15 Mm³, and a scenario written on the commercial figure alone is short by a quarter of the country's wood. The default is the SNBC 3's 60 Mm³/y from 2030. 75 is the top of the IGN–FCBA's B3 case and close to the 71 of the two harvest-heavy ADEME scenarios. 40 is low but not unpublished: the Fern–Canopée–Amis de la Terre report takes the harvest down to 30 Mm³/y by 2050 in its most extensive scenario — on a basis that counts branches and dead wood, so about half of what it counts today — and it is also the report that prices that scenario: a supply crisis, sawmill closures and rising imports, which is why its own authors recommend holding 60. Moving this lever moves the land sink by 1.87 tCO₂ per cubic metre — 2.0 in the standing forest, less the 0.13 that cubic metre would have put into wood products — which is the point: until this module existed the game could burn as much wood as it liked and the forest never noticed. What the model does not do is replace the wood that is not cut: no import, no concrete in its place. The controversy table says why that is contested, and since 0.26.0 the scoreboard charges a scenario that leans on its land for more than the SNBC 3's own −23 MtCO₂e.

• IGN, Mémento de l'inventaire forestier 2024, pp. 36–37 — removals 53.1 ± 2.9 Mm³/y (2014–2022)
• SNBC 3 — 60 Mm³/y from 2030 (63 in the November 2024 draft)
• IGN–FCBA (2024), Projections des disponibilités en bois — B1/B2/B3 at 53 / 63 / 75 Mm³/y
• du Bus de Warnaffe G. & Angerand S. (2020), Gestion forestière et changement climatique, Fern–Canopée–Amis de la Terre, pp. 5–6 and 31 — harvests of 30 to 95 Mm³/y in 2050 from 60 today, and the cost of the low one
• France Stratégie (July 2023), Note d'analyse n° 124, Vers une planification de la filière forêt-bois, p. 12 — the trade-deficit cost of the lower-harvest strategy

Flagged below 53 Mm³/y. Below 53 the scenario cuts less wood than IGN measures France cutting today — 53.1 Mm³, informal firewood included — and the wood no longer cut is replaced by nothing this model counts. Every cubic metre spared books 1.87 tCO₂ of absorption: fifteen million fewer than the SNBC 3 asks for is twenty-eight megatonnes a year. The French literature never treats that as free — France Stratégie names the cost of the lower-harvest strategy as a wider trade deficit, importing the chips, sawn wood, pulp, panels and furniture the country still wants — and the scoreboard now charges it, on the land-sink line.

Only sawn timber and panels hold their carbon for decades — fifty years for structural framing, thirty for flooring, twenty-five for panels — while pulp and packaging give it back within about seven. The SNBC 3 draft moves sawn from 9.5% to 12% of the harvest and panels from 13% to 18%, i.e. 22.5% to 30%, and expects the wood-products pool to go from roughly zero today to at least 3 MtCO₂e/y in 2030. The IGN–FCBA study's most useful finding is that reallocating the marginal cubic metre from energy to material is worth about as much as increasing the harvest at all. This lever is that finding, and it trades against the wood supply: at a fixed harvest, every point that goes into a long-lived product is a point that does not go into a boiler.

• Projet de SNBC 3, 31 October 2024, pp. 88–96 — sawn 9.5→12%, panels 13→18%, wood-products sink ≥3 MtCO₂e/y in 2030
• IGN–FCBA (2024), Projections des disponibilités en bois, p. 67 — allocation is worth about as much as management

Not shown in this edition. This country's package hides the control; the model reads the lever at the default in the next column.

Declared, hidden and provably inert. France's land_class carries no drained organic soil (see the table's own why), so every term this lever enters is a multiplication by an area of zero and a test asserts that its whole range moves no output at all. It is hidden rather than drawn because a slider that cannot change a number is a lie about the model rather than a gap in it. The day Citepa publishes an organic-soil area by land-cover class, this becomes a real French lever and its bounds become a real French argument.

Three positions, from the IGN–FCBA climate cases: by 2050 production falls 1%, 12% or 25% and mortality rises to 1.1, 1.4 or 1.8 times today's. Mortality has already doubled since 2005–2013 — spruce, chestnut and ash are dying now — so C2 is the central case rather than the pessimistic one. This control is what makes the −62 to +5 MtCO₂e band the Haut Conseil pour le Climat puts around the SNBC 3's own sink reachable in the game without inventing a curve. At C3 with a hard harvest the French forest becomes a net source, which is not a modelling artefact: it is what the published severe case says, and the module reports it rather than clamping it away. The player chooses it, which is uncomfortable — nobody chooses their climate — and the alternative was to pick one for them.

• IGN–FCBA (2024), Projections des disponibilités en bois, p. 22 — climate cases C1/C2/C3
• Haut Conseil pour le Climat, Avis sur le projet de SNBC 3 (2026), §3 — LULUCF +5 / −23 / −62 MtCO₂e excluding forest soils

Beef, pork and sheep meat: 20.8 + 30.6 + 2.1 = 53.5 kgec a head in 2024, against 85.0 kgec of all meat. The default of 40 is a quarter below the observed diet and sits between the PNNS 4 recommendation and the −40% every study that reaches −40% greenhouse gases lands on; the minimum is the EAT-Lancet reference diet's order of magnitude, and the maximum is above anything France has eaten since 2004. The SNBC 3 does not publish a diet in kilograms — it says "PNNS-conform" and "limited", and cuts imported meat first — so the default here is the module's reading of a strategy that declined to give a number, not the strategy's own figure. That is worth knowing before quoting it.

• Agreste/FranceAgriMer, Synthèses n°443 — consommation de viande 2024, table 1
• INRAE, Propositions d'hypothèses pour le scénario AMS de la SNBC 3 (April 2023), p. 9 — −40% all meats
• SNBC 3 (July 2026), Partie 1, pp. 114–128 — a PNNS-conform diet, no tonnage published

Flagged below 20 kgec/cap/y. 53.5 kgec is the observed 2024 diet, and every study that reaches −40% greenhouse gases lands around 32. Below 20 the scenario is at the EAT-Lancet reference diet's order of magnitude: a fall of more than three fifths in a single generation, and a change in what a country eats that nothing in this model brings about, prices or even makes anyone argue for.

Poultry is separated from red meat because it is the one meat whose consumption rises — 2.2% a year, from a quarter of French meat to a third in twenty years — and because every diet scenario cuts it least. A model that moved all meat together would hide the substitution that is actually happening.

• Agreste/FranceAgriMer, Synthèses n°443 — 2 133 kt of poultry, +7.1% on 2023

An index rather than a quantity, because milk equivalents are published on three incompatible bases and none of them survives a conversion cleanly. The default of 90 is a tenth below today, well short of INRAE's −30%: dairy is the lever that fights back hardest, because cutting milk cuts the dairy herd and the dairy herd supplies two fifths of the beef, so the suckler herd grows to meet a beef demand that has not moved. Moving this slider alone shows that coupling better than any chart.

• INRAE, hypothèses AMS SNBC 3 (April 2023), p. 9 — dairy −30%
• Solagro, Afterres2050 version 2016 — dairy 235 → 122 g/day

It is a small lever, and the arithmetic says why. The share it acts on is the edible fraction of the food supply — 3.8 Mt of edible waste on a supply of order 55 Mt, so 7% — and halving that removes about three and a half per cent of the demand. The thirty per cent everyone quotes is ADEME's share of the food chain's losses that occur at consumption, which is a different quantity and about four times larger; putting it here would make food waste look like the biggest lever in this module rather than one of the smallest. The default is zero and not the SNBC 3's −50%, because the strategy's target is set against 2015 and on the whole chain, and reading it onto an edible-waste share of a 2024 supply would be arithmetic the strategy did not do.

• SDES, Les déchets alimentaires en France et dans l'UE en 2023 — 9.7 Mt of food waste, 3.8 Mt of it edible
• ADEME (2016), Pertes et gaspillages alimentaires, p. 4 — 10 Mt/y, 33% at consumption

A rule, and the only lever in this module that no source sets a 2050 level for. France exports two fifths of its milk while importing a third of the dairy it eats, and exports about a fifth of its pork; without this lever a French diet change would move the French herd one for one, which is simply wrong. With it, a player can eat less meat and keep the herd, or keep eating and stop exporting, and see that these are different decisions with different answers. The volumes are indexed rather than the shares, because a share runs away as it approaches one: at an export share of 90% a five-point move doubles the herd. The bounds are the game's, not a study's, and nothing published says what France should export in 2050.

• CNIEL, rapport annuel 2024, p. 35 — 40% of French milk production exported
• Agreste, Synthèses n°440 Bovins (June 2025) — 25.5% of beef consumption imported

France exports 26.9 Mt of cereals a year on a harvest of 60.9 Mt, the average of the five campaigns 2020/21 to 2024/25 — 44% of what its fields grow, half of it soft wheat, and the largest use of its arable land after feeding the herd. A domestic diet change therefore moves exports before it moves fields, and a crop block that ignored the export position would find France with idle hectares the day it eats less bread. The lever is indexed on the base-year volume, exactly as livestockExport is, and the default holds it: no published scenario sets a 2050 export level, INRAE only notes that a smaller herd and less ethanol free cereals for export. The maximum is half as much again.

• FranceAgriMer, Fiche filière céréales 2026 — averages 2020/21–2024/25: 60.9 Mt produced on 8.9 Mha, 26.9 Mt exported, 54% of it soft wheat
• INRAE, hypothèses AMS SNBC 3 (April 2023), p. 11 — 'augmentation capacité export céréales, car moindre utilisation pour élevages et pour bioéthanol'

The SNBC 3's −54% against 2020 is 55% of the 2024 delivery, and the reference delivers exactly that — but since stage E the lever is the dose on the hectares that stay conventional, because the organic share takes its hectares out of the mineral dose on its own. INRAE's decomposition of the strategy's cut is the reason: of the −944 kt N it books, 330 come from extending organic farming to a quarter of the area, and reading the −54% onto a dose and moving the organic lever would have counted those 330 twice. So the default is 70% on the conventional hectares, which with organic at 25% delivers 55% of 2024 in all; INRAE's own proposal is −46%, TYFA goes to zero on a fully organic Europe, and the maximum is a little above the 2020 level, which 2024 has already fallen below. It is the lever with the longest reach in the module: it sets the soil N₂O and the urea and liming CO₂ in agriculture, and the ammonia the French industry chain has to make, and the hydrogen that ammonia draws from the hydrogen mix. Halving French nitrogen is worth about five megatonnes in agriculture and about a million tonnes of ammonia in industry at the same time, and until this module those were two unconnected numbers. Since 0.24.0 it also sets a yield. Below 90% of the 2024 dose (n_yield_plateau) the conventional hectares lose yield along the GRAFS hyperbola — they keep 0.90 of it at the reference 70% and 0.74 at the floor of 40% — and the crop block asks for the arable land that loss costs. Above 90% a heavier dose buys nothing, which is why the maximum of 110% only adds nitrous oxide.

• SNBC 3 (July 2026), Partie 1, p. 128 — mineral nitrogen −30% by 2030, −54% by 2050 against 2020
• INRAE, hypothèses AMS SNBC 3 (April 2023) — −944 kt N, −46%

Flagged below 50%. The SNBC 3's own cut — −54% against 2020 — is 55% of the 2024 delivery, and the reference books it already; INRAE's own proposal is milder still, at −46%. Below 50% of the base-year dose on the hectares that stay conventional the scenario is past every published French trajectory. Only TYFA goes lower, and it does so on a fully organic Europe, with the yields and the diet that implies.

The SNBC 3 multiplies the legume area by two by 2030 and by nearly three by 2050, from one million hectares to 2.7, and the default is that target. INRAE books a credit of 128 kt of mineral nitrogen for an increase of 1.7 Mha, which is the coefficient this module reads, linearly over that span. The hectares sit inside arable land and move no class of the land account: a legume is a crop in a rotation, not a change of land use. What they do compete for is the same arable land the energy crops of stage C will want, and arable_committed is where that competition is reported.

• SNBC 3 (July 2026), Partie 1, p. 128 — legumes 1 → 2 → 2.7 Mha
• INRAE, hypothèses AMS SNBC 3 (April 2023) — 128 kt N credited to +1.7 Mha of legumes

The SNBC 3 takes organic farming from 5.6% of the field-crop area in 2024 to 21% in 2030 and 25% in 2050, and the default is that target. The lever is stated on the arable area rather than on the whole farmland, because that is where both of its effects sit: the 10% of the agricultural area the organic agency publishes for 2024 — 2.7 Mha — is mostly grassland, which takes little mineral nitrogen and has no yield gap worth modelling. The maximum, 50%, is Afterres2050's organic share, and ADEME's S1 goes to 70%. Two consequences, and the second is the one the module was missing. An organic hectare takes no mineral nitrogen, so the mineral dose falls in proportion — the INRAE hypotheses for the strategy book −330 kt N for the same extension, and this lever gives −360 at the base-year dose. And an organic hectare yields about two thirds of a conventional one, so the same plates, the same herd and the same exports need more land: at 25% the arable area the country needs rises by 7%, and at 100% by half. The land account does not resolve that tension; it reports it as arable_headroom, negative where the fields the diet needs do not exist.

• SNBC 3 (July 2026), Partie 1, p. 127 — organic farming 21% of the field-crop area in 2030 and 25% in 2050, against 5.6% in 2024
• agriculture.gouv.fr, infographie 'L'agriculture biologique' (Agence Bio, chiffres 2024) — 2.7 Mha in organic production, 10% of the agricultural area
• INRAE, hypothèses AMS SNBC 3 (April 2023), p. 11 — 'extension agriculture biologique sur 25% de la SAU : −330 kt N'
• Solagro, Afterres2050 (2016) — organic 45% plus integrated 45% of the area; ADEME Transition(s) 2050, S1 — organic 70%

The SNBC 3 puts lipid-enriched rations — linseed, rapeseed — on 82% of housed cattle by 2050, at −14% of enteric methane where they are fed, after Pellerin's 2013 assessment. That is the default, so this lever is one of the ones reachable only downwards from the reference. The additive that would do better is not in it. 3-NOP cuts enteric methane by 20–35% in dairy cows in the trials, but it is fed only while the animals are housed, it is not in the SNBC 3, and putting its number here would be claiming a strategy France has not adopted. Raising the maximum to 100% is as far as this module goes.

• SNBC 3 (July 2026), Partie 1, p. 128 — lipid rations on 82% of housed cattle, −5% herd-wide
• Réussir Lait on EFSA and INRAE Theix — 3-NOP −20 to −35%, not in the SNBC 3

The default is zero, and it is not the SNBC 3's 80%. The strategy's figure is the maximum here and is reachable, and it is worth about three megatonnes; what it is not is a number the reference scenario should book. The abatement rests on a split between enteric and manure methane that Citepa publishes only as a total — the per-species manure_ch4_share in the livestock table is an assumption, not a measurement — and booking three megatonnes on an unpublished split would have closed most of this module's gap to the SNBC's own 2050 agriculture figure with a coefficient nobody can check. The gap is information; see the changelog. What the lever will also do, from stage C, is produce biogas. It is declared here because it belongs to the manure block, and its supply side is the next stage's.

• SNBC 3 (July 2026), Partie 1, p. 128 — manure methanised 22% (2030) → 80% (2050)
• Citepa, Rapport Secten éd. 2026, p. 451 — 1 412 farm digesters in 2024; enteric and manure methane published together

The SNBC 3 takes fossil fuel on farms to zero by 2050, and the default is that. It is worth 10.73 MtCO₂e — agriculture 10.33 plus forestry 0.41 — which is a seventh of the sector and the single largest thing a French farm can stop doing. What it does not say is what runs the tractors afterwards, and neither does the strategy. This module therefore books the emissions and not the energy: the roughly forty terawatt-hours behind the line are outside the model's carrier pools in both directions, and that is stated in the farm_fuel_emissions equation rather than papered over with an invented electricity demand.

• Citepa, SECTEN 2026 — engines, motors and boilers, agriculture 10.33 and forestry 0.41 MtCO₂e in 2024
• SNBC 3 (July 2026), Partie 1, p. 128 — fossil fuel on farms at zero in 2050

France Fertilisants puts French production at 34% of the nitrogen French farms use, with 24% from other EU countries and 42% from third countries — mostly urea, UAN and DAP from Russia, the United States, Egypt, Algeria and Trinidad. That is the default, and it replaces the free-standing ammonia tonnage the model used to carry. At the base year's 1 817 kt of nitrogen and 34%, the derived tonnage is 900 kt, which is what the retired slider asserted — a cross-check rather than a fit, because the share comes from the fertiliser industry and the nitrogen from Citepa, and neither was chosen to land there. Moving this lever is a sovereignty argument with an energy bill attached: making all of France's nitrogen at home multiplies the ammonia and the hydrogen it draws by three.

• France Fertilisants / UNIFA — la fertilisation azotée en France: 34% national, 24% EU, 42% third countries

2.5 Mha is the energy directorate's own figure (2.55), and it is the default because it is the one a national scenario has actually written down. The mission that reviewed the biomass potential recommends "about 3", which is the maximum here; INRAE's expertise says 4 Mha of French spring cropping could carry a cover crop at all, which is cive_land_ceiling and is deliberately not reachable by the slider. This is the largest single term in the biogas supply — 2.5 Mha at 6 t DM and 2.8 MWh/t is 42 TWh, half of everything the module produces — and it is also the one an agronomist will argue with first. A cover crop takes no food land, but it needs mineral nitrogen in most years and this model does not charge it; the interannual variation of the yield is ±5 Mt DM on 15, which is a third of the term.

• IGEDD/CGE/CGAAER 2023, p. 22–23 — 2,5 Mha (DGEC 2,55), la mission recommandant ~3 Mha ; INRAE 4 Mha
• Note de recherche du dépôt, docs/agriculture/bioenergy.md §4.2 — CIVE

16% is what an INRA/ADEME study finds can be exported without a soil carbon loss — 55% of the surplus straw, which is 4.1 Mt of fresh matter. Solagro caps at 20%; a 2025 agronomic study finds more than 30% is possible at département scale while keeping the fifty-year soil-carbon loss under 2.5%, which is the maximum here. The ceiling is soil carbon and nothing else: a tonne of straw left on the field returns 400 kg of carbon and 6 kg of nitrogen to it. Push the slider and the land account does not charge you for the carbon, because the inventory's cropland pool is calibrated on the base year's practice — a named limitation, and the reason the maximum is 30 and not 50.

• INRA/ADEME (AGRICE), Prélèvement de paille — 4,1 Mt exportables, 55 % du surplus
• Agro-Transfert, Exporter des pailles sans risque pour l'état organique des sols, 2025 — >30 % pour <2,5 % de perte à 50 ans
• Solagro, Quelles biomasses pour la transition énergétique ?, oct. 2024, p. 20

Flagged above 25%. 16% is what the INRA/ADEME work finds can leave the field without a soil-carbon loss, and Solagro caps at 20%. A tonne of straw carried away takes 400 kg of carbon and 6 kg of nitrogen with it — and the cropland pool here is calibrated on the base year's practice, so the land account does not charge the loss back. Past 25% the biogas is real and the soil carbon it costs is invisible.

0.618 Mha today, and the default is that rounded to the slider's step: 218 kha of ethanol crops (wheat 115.8, maize 75.2, sugar beet 27.2, FranceAgriMer 2026) and about 400 kha of rapeseed and sunflower for FAME. The 400 kha is the weakest number in this block — it is a FranceAgriMer estimate quoted by the trade press and was not verified at source — and it carries 6.8 of the 11.7 TWh the term produces. The maximum is the mission's own proposal: +1.1 Mha of rapeseed and beet for +19 TWh. Note what the slider does not say — the "about one million hectares" often quoted for French biofuels is the European footprint of French consumption, and half of what France burns is imported as fuel or as feedstock. French land in French biofuels is 0.6 Mha for 11.7 TWh.

• FranceAgriMer, Fiche filière bioéthanol 2026, p. 1 — 218 kha : blé 115,8, maïs 75,2, betterave 27,2
• IGEDD/CGE/CGAAER 2023, annexe 10 — +1,1 Mha (colza + betterave) pour +19 TWh
• FranceAgriMer via lafranceagricole.fr — ~400 kha de colza et tournesol pour l'EMAG. NON VÉRIFIÉ À LA SOURCE ; l'ONRB de FranceAgriMer le fermerait

Not shown in this edition. This country's package hides the control; the model reads the lever at the default in the next column.

Zero, with a range of zero, and hidden. French methanisation runs on manure, waste and cultures intermédiaires à vocation énergétique, which is what civeArea is: the SNBC 3 and the biomass strategy both rule out a dedicated main crop, and the maize grown in France is grown for silage and for grain. The lever exists because Germany's does, and it is fixed at its declared value rather than merely hidden so that its inertness is arithmetic rather than a promise — the minimum and the maximum are the same number, so no run of the model can move it.

• Décret n° 2016-929 du 7 juillet 2016 — cultures alimentaires ou principales limitées à 15 % du tonnage brut d'intrants d'une installation de méthanisation
• Projet de SNBC 3, 31 October 2024 — methanisation on manure, waste and cover crops, with no dedicated main crop

France imported about 19 TWh of finished biofuel and 11 TWh of feedstock in 2023; the ecological planning secretariat's own 2030 balance keeps 12 to 20 TWh of liquid imports. 20 is the middle of that and the default; 40 is roughly today's gross import position, which a 2050 that has not reduced its liquid demand would still need. The bounds are a rule dressed as data and the model says so. No published French study sets a 2050 import allowance for liquid biofuel, so what anchors this slider is today's trade position and one 2030 figure. Wood imports (5–10 TWh) are left at zero because they are small, and there is no biomethane import line in any French study at all. What it buys is worth stating plainly: the good band is the domestic supply, so importing fuel moves the warning line and never the target. A scenario that meets its liquid demand on imports is amber by construction.

• SGPE, Bouclage biomasse, juillet 2024, p. 18 — 34 TWh de matière première (11 importés) → 35 produits (14 exportés) → 40 utilisés (19 importés) ; 12–20 TWh d'importations liquides en 2030
• IEA Bioenergy, Implementation of bioenergy in France, 2024 — importations nettes : biodiesel 48 % de l'usage, éthanol 30 %

Flagged above 30 TWh/y. No French study sets a 2050 import allowance for liquid biofuel: what anchors this slider is today's trade position and one 2030 balance, which keeps 12 to 20 TWh. Past 30 the scenario meets its liquid demand on fuel grown somewhere else — on land this account does not hold, under a sustainability regime it does not check, and with an emission the game perimeter never sees.

A share of the service, not of the energy. Switching a gas water heater for an electric one does not move the same number of kilowatt-hours: the electric route delivers the same hot water from less energy, and the model converts through the two efficiencies rather than shifting the energy across unchanged. 68% is where France is today — of the service, which is why it is not the 47% an energy split would give: a kilowatt-hour of electricity delivers more hot water than a kilowatt-hour of gas. At that default the model reproduces the observed energy exactly. The 26 TWh of gas, oil and LPG behind the remaining third is one of the two things standing between a maximal scenario and a winnable game. One approximation: a single share is applied to residential and tertiary alike, where today they sit at 75% and 51%. The national total is exact; the split between the two segments is not.

Same treatment, and the efficiency gap is much wider here: a gas hob puts about 40% of its energy into the pan and an induction plate about 84%, so electrifying cooking roughly halves the energy it takes. 61% of the service is electric today, which reproduces the observed energy exactly.

Cooling is the one building usage certain to grow, and the model cannot score it properly: it makes a summer peak, and the only peak constraint here is a winter one. The number is carried and the asymmetry is stated. The 24 TWh is the building_usage table's own two cooling rows, residential plus tertiary.

Lighting, appliances, screens and servers. It is the lever that has historically delivered — French specific consumption has been roughly flat for a decade while the equipment count rose — and here it is set against the growth lever below, which is the whole point of having both.

The counterweight to appliance efficiency. Data centres and AI are the part of this that is growing fastest and the part the model is least able to source, so it is left as an explicit assumption rather than given a trajectory it cannot defend.

RTE keeps a few GW of combustion capacity in every 2050 scenario, for the windless fortnight that no amount of storage covers. What it burns is a choice: methane, which draws on the same biomethane everything else wants, or hydrogen, which draws on electricity instead and emits nothing at the stack. Hydrogen is much the dearer of the two, and the model charges it: the electrolytic price the industry module already computes, against a methane price. What it does not do is add the electrolysis back into the electricity the mix has to serve — that would be a fixed point the compiler cannot express, since demand sets the mix and the mix would set demand. The extra electricity is reported instead of hidden.

The model assumed this for every tonne of hydrogen until v0.12.0, which was a strong assumption wearing no clothes: 87 TWh of electricity for hydrogen, and no way to ask what a reformer would cost instead.

A reformer burns and reforms methane. In this model 2050 methane is biomethane, so the colour of the hydrogen depends on the colour of the gas — and it draws on the same biomethane pool as everything else, which is the trade-off worth seeing.

ATR concentrates the CO2 in one stream, which is why it captures at 94% where a reformer with post-combustion capture struggles past 60%. On biomethane this goes negative, and that is not a trick of the accounting: the carbon came out of the air last season and is being put underground. It is also the single most contested line in the model — see the controversy tab — because it makes a scenario's arithmetic depend on a biomass supply chain the model does not represent.

Which of RTE's six 2050 mixes the scenario is built on, from M0 at 100% renewable to N03 at about half nuclear. It selects a set of shares, not a quantity: the mix is scaled to whatever electricity the rest of the model turns out to need, so choosing a scenario here answers "with what" and never "how much".

• RTE, Futurs énergétiques 2050 — scénarios de mix 2050

The chemistry choice, and the sharpest trade-off in the account. LFP carries almost no cobalt (7 grams per MWh against 27 kg) and a quarter of the nickel, but 4.4 times the lithium — 490 kg per MWh against 111. There is no chemistry that is cheap in every metal at once, which is the point of putting it on a slider. Zero by default because the source scenario's 2050 reference is NMC 811.

• POMMES-INDUSTRY, France 2050 — finance_rate

A household and an industrial investor do not face the same cost of capital. Moving this rate from 4% to 8% raises the building indicator by about a third with no physical change at all, which is why it is a lever and not a hidden constant.

• POMMES-INDUSTRY, carbon.csv — linear to 150 €/tCO₂ in 2050

• POMMES-INDUSTRY, import_hourly.csv, France 2050

No primary publication has been secured for this figure. It is exposed as a slider rather than hidden as a constant so the uncertainty is testable. Securing the CSTB renovation-gesture database is the single change that would most improve the building cost module.

• ADEME renovation cost orders of magnitude; CSTB gesture database derived from Batiprix 2022 (used by OptoBat, not public)

A 2050 pump price for biofuel and e-fuel, taxes included. No source has been secured; the aviation module now prices synthetic fuel bottom-up and is the better anchor.

• No primary source secured — see the aviation fuel-cost table for a bottom-up range

The passenger sufficiency control. At 100% it removes 45% of passenger travel and takes air traffic from the workbook's implicit 0%/y to -1.5%/y — the two demand assumptions the transport module is most sensitive to, moved together because a scenario that flies as much as today while driving 45% less is not a coherent story about sufficiency.

Moves the 20% of 2020 car demand still served by a liquid- or gas-fuelled car in the reference onto the electric fleet. The rail transfer is left where it is: electrifying the fleet and shifting trips off it are two different decisions, and folding them together would hide which one paid.

The reference already leaves only 10% of road freight thermal, so this is a small lever by construction, and that is the lesson: on the workbook's own trajectory the remaining road-freight emissions are not where the tonnes are. Hydrogen trucks are left alone, because arbitrating between hydrogen and battery is a detailed-view argument, not a coarse one.

Modal shift away from aviation, which is where the transport module's residual emissions concentrate once the fleet is electric. Distinct from "travel less": this one moves the same journeys onto another mode rather than removing them.

The building sufficiency control, in the unit of the lever it drives because it drives only that one. It starts at the reference 5% rather than at zero: the simple view offers effort beyond the reference scenario, never less than it.

As with sufficiency, this is the stock-performance lever shown in its own unit and bounded below by the reference scenario. 65% across the whole stock is the upper end the workbook contemplates, not a technical limit.

The materials side of the building question, and the one control in the simple view whose whole interest is how far it cannot reach. At 100% the country builds 8 Mm² of housing and 8 of everything else a year instead of 20.4 and 19.9 — the order of magnitude the official housing-need study reaches for the 2040s — and frames 60% of it in timber instead of 12%. Together those take roughly a quarter off national cement demand. They take far less off steel, because new buildings are about a ninth of it; and they cannot touch the third of cement that is roads, buried networks and bridges, nor the third that nobody has attributed. Two ideas in one control is a deliberate departure. Building less and building in timber are different decisions, and the detailed view keeps them apart; here they are bundled because they are the same material choice seen from a distance, and because a simple view that separated them would have spent two of its dozen controls on one question.

Electrification of heating, with the two consequences that make it a real choice rather than a free win. The wood boilers go with the gas ones, because 95% electric plus 46 TWh of wood would allocate more heat than the stock needs; the resistance heaters go into air-water heat pumps, because electrifying on resistance is what makes the winter peak unmanageable. The peak still rises sharply, and that is the point of the control. Pushing sufficiency and renovation at the same time shrinks the heat need under a fixed 95% share, so the building readout may report over-allocated heat; the model reports it rather than absorbing it.

Hot water and cooking are about as much energy again as space heating, and neither is fully electric today. Their observed electric shares are the starting points, 100% the end.

The industrial sufficiency control. Freight demand is driven here rather than in transport on purpose: freight is what material consumption looks like on a road, and a scenario that halves plastic and cement demand while moving the same tonne-kilometres is not consistent. Steel goes from +30% to -40% against 2020, which is the widest swing any single number in the game commands. The output of the rest of industry is not driven, because its scale starts at today's output and has nowhere lower to go.

Every route change the five value chains offer, moved together: H-DRI steel, the CO2 + H2 olefin route and the biogenic carbon it uses, heat pumps for food-industry steam and direct heat, the lowest clinker ratio in range, and capture on what is left. The processes of the rest of industry are not driven, because the reference already sits at 100% of them.

Efficiency rather than sufficiency or fuel switching: the same product, less energy. At 100% it claims the whole potential RTE identifies and recovers all of the recoverable waste heat, neither of which is costless or instantaneous — the cost panel and the annex say what that means.

The demand end of the food chain, moved as one plate. At 100% red meat falls from 40 to 20 kgec/cap/y — below INRAE's −40% and above ADEME's S1 divide-by-three — poultry from 28 to 18, dairy to 70% of the base year and the edible waste by the SNBC 3's own half. Exports are deliberately left alone: what a country sells is a separate argument from what it eats, and folding them together would let a diet lever cut a herd that is producing for somebody else's plate. The lesson is in the coupling rather than in the total. Two fifths of French beef is a by-product of the dairy herd, so cutting the milk makes the suckler herd grow to meet a beef demand that has not moved; only moving both together shrinks the cattle. A player who moves this control and watches the grassland released is watching the land account answer a food question, which is what the module is for.

The sink side of the land account. At 100% the forest expands at the 90 kha/y IGN's inventory measures rather than the 15 the SNBC 3 plans, the harvest falls from 60 to 45 Mm³/y, the long-lived share of what is still cut rises from 30 to 35%, and artificialisation stops entirely — the Climat et Résilience law's 2050 destination, on this account's own measure. It is the control that most obviously costs something, and that is the point: the forest identity is k · (P·A − M·A − H), so every cubic metre not cut deepens the sink and leaves the boiler. Pushed to the top this control takes about a fifth off the wood the scoreboard scores — 127 TWh to 102 — and a scenario that has also electrified its heating will not notice while one that leaned on wood will. The climate case the forest lives through is not driven: it is a scenario choice rather than an effort, it moves the answer further than any of these four, and it stays a fine lever the simple view draws on its own.

The field end of the farm. At 100% the mineral dose on the conventional hectares falls to 50% of the base year and half the arable land goes organic, which between them deliver 26% of the 2024 nitrogen — well below the SNBC 3's own −54% and short of TYFA's zero — the legumes reach the 3.0 Mha the rotation studies stop at, and the whole 17.3 MtCO₂/y of soil-carbon practice INRAE itemises is taken. Four consequences are worth watching rather than assuming. The organic half yields two thirds of what it replaces, and since 0.24.0 the conventional half, at half the 2024 dose, keeps 0.79 of its yield — it is below the plateau — so this control costs arable land: the fields the same plates, herd and exports need grow by more than a third, and the crop block reports the shortfall, 4.2 Mha at 100%, against the land account rather than closing it — fertilising less is not free of land, and the page says by how much. Mineral nitrogen is also an industrial decision here: the same tonnage sets the ammonia the chain has to make, so fertilising less is a hydrogen saving as well as a nitrous-oxide one. More legumes raise the nitrogen balance while lowering the emissions — a legume hectare fixes more nitrogen than the mineral fertiliser its credit replaces, and only the difference between the two emission factors makes the net move downwards, so the balance is not a proxy for the tonnes. And the soil-carbon target is a thirty-year rate on a stock that saturates: the practice has to be kept up after 2050 for the carbon to stay where this account puts it.

The supply side of the biomass the rest of the game spends. At 100% the winter cover crops reach 3.0 Mha, the straw taken off the field reaches 30% — Agro-Transfert's ceiling for less than 2.5% of soil carbon lost — and the land growing fuel reaches 1.70 Mha, nearly three times today's and the IGEDD mission's own upper case. Between them they take the methane supply from 70 to about 86 TWh and the domestic liquid supply from 24 to 52. Two things this control does not do. It does not send manure to a digester: manureMethanised sits at 0 in the reference because the abatement it books rests on an enteric/manure split no French inventory publishes, and burying that argument inside an effort scale would be the wrong place for it. And the fuel crops it plants add no mineral nitrogen, because the dose is an intensity on an arable area held at the base year's: a fuel hectare displaces a food hectare that was already fertilised. The crop block counts the displaced food — the fuel hectare is arable land the plates, the herd and the exports can no longer use, and the headroom says by how much — but the fuel hectare itself adds no nitrogen and no nitrous oxide. That is the module's largest structural simplification and it sits directly under this control.

The workbook converts electricity to hydrogen at 60%. POMMES-INDUSTRY uses 45 MWh of electricity per tonne of hydrogen, which is 74%. The model keeps the workbook value so the cost layer and the electricity indicator describe the same hydrogen; hydrogen is therefore about 40% more expensive here than a POMMES-native calculation gives.

• Teaching workbook, "General hypotheses" sheet E11

Electricity in, liquid e-fuel out, across the whole chain: electrolysis, CO2 supply, synthesis and upgrading. Concawe's techno-economic assessment gives 38% when the carbon comes from direct air capture and 44% when it comes from a concentrated industrial source, and 0.4 sits between the two. It is a whole-barrel figure, not a kerosene-only one: Fischer-Tropsch sends roughly a fifth of its liquids to gasoline and diesel, so the electricity behind one tonne of jet fuel alone is higher — about 38 MWh, or 3.2 kWh per kWh burnt. Use this parameter for total liquid demand, which is what the equations do, and do not quote it as an aviation figure.

• Concawe 2022, E-Fuels: a techno-economic assessment of European domestic production and imports towards 2050 (38% DAC, 44% CCU)
• Boigontier & Girard 2024, Sustainable Aviation Fuels (SAFs), MINES Paris - PSL / Zenon Research
• Teaching workbook, "General hypotheses" sheet E10

• Teaching workbook, "Synthesis" sheet S41 and "Industry" sheet D37

A gas or oil water heater, standing losses included.

• ADEME orders of magnitude for domestic hot water production

Above one because it is not all resistance: a 2050 electric water heater stock is a mix of resistance tanks and heat-pump water heaters, the latter at a COP near 3. Two is the blend assumed here and it is an assumption, not a measurement — the honest range runs from 1.0 if nothing changes to near 3 if the stock is heat pumps. It sets how much electricity electrifying hot water actually costs, so it is worth arguing about.

• ADEME, chauffe-eau thermodynamique — COP de 2,5 à 3,5 selon l'installation

• Standard hob efficiencies; a gas burner loses most of its heat around the pan

Induction. The gap with gas is the widest of any usage in the model.

• Standard induction hob efficiency

The carbon actually in the molecule, released whether it is burned or reformed. Distinct from efGas, which is 25 gCO₂/kWh in 2050 because the model's methane is biomethane and the convention books its carbon against the digester feedstock rather than the flame. Both numbers are needed and they answer different questions. efGas answers "what does burning this count as?"; this one answers "how much carbon is there to capture?" A capture plant removes molecules, not conventions.

• Base Carbone ADEME — combustion du gaz naturel, facteur direct

• Teaching workbook, "Synthesis" sheet S42

• Teaching workbook, "Synthesis" sheet S44

• Teaching workbook, "Synthesis" sheet S43

Coking coal, 94.6 kgCO₂ per GJ, converted at 3.6 GJ/MWh. The workbook declares coal at 0 gCO₂/kWh in its factor table and then charges it at the hydrogen factor, 66.7 gCO₂/kWh, which is neither. This model uses the published factor and removes the double count that resulted — see steel_bf_process_residual.

• IPCC 2006 Guidelines for National Greenhouse Gas Inventories, vol. 2 table 1.4 — coking coal, 94.6 kgCO₂/GJ

Crude steel by the blast-furnace route, France 2020, split by the H-DRI lever in 2050. worldsteel publishes France at 11.6 Mt that year, 67.5% oxygen converter and 32.5% electric furnace; Eurofer's more precise total is 11 596 kt. It was 9 900 kt, and that figure was 2015, not 2020. France made 9 825 kt by this route in 2015, out of 14 984 kt — the last year the country reached fifteen million tonnes, and the base year of ADEME's sectoral plan for steel. Since then: 10 061 kt in 2019, 7 830 in 2020, 9 280 in 2021, and about 6 300 in 2025, with one of Fos's two blast furnaces idle from October 2023 to mid-2026. On the old figure, steelGrowth's "+30% on 2020" made 19.5 Mt of French steel in 2050, which is +68% on 2020; it now makes 15.1 Mt, the country's 2015 level. The limitation. 2020 is a Covid trough, 20% below 2019. It is used because it is this model's base year, not because it is typical, and the route split after 2019 is published free only as a rounded percentage, so the figure carries about ±50 kt of rounding.

• worldsteel, World Steel in Figures 2021, p. 10 — France 2020: 11.6 Mt, 67.5% oxygen / 32.5% electric
• Eurofer, European Steel in Figures 2024, p. 16 — France 2020: 11 596 kt
• worldsteel, Steel Statistical Yearbook 2017, tables 7–8 — France 2015: 9 825 kt oxygen, 5 159 kt electric
• See docs/industry/steel_and_cement_base.md for the series 2010–2025 and the fleet

Crude steel from scrap in an electric furnace, France 2020: 32.5% of 11.6 Mt. It was 5 100 kt, which was 2015 (5 159 kt observed). The electric route is the more resilient of the two: between 2015 and 2025 it fell 30% against 36% for the blast furnace, so its share rose from 34% to 36.4% even as its tonnage fell. A3M counts fourteen electric steelmaking sites in France making about 4 Mt a year.

• worldsteel, World Steel in Figures 2021, p. 10 — France 2020: 11.6 Mt, 32.5% electric
• A3M and Federrec (September 2026), Pacte Acier Vert — 14 electric sites, about 4 Mt

The total direct intensity of the integrated route — every source on site, not a process term. It was inherited from the teaching workbook as "the blast furnace's process figure", and under that name it had been counted on top of the coal it mostly is. Re-sourced in 0.27.0, it holds: the IPCC Tier 1 defaults summed over the chain give 1.87 tCO₂/t (1.46 for the converter, which already includes the blast furnace, plus 0.56 per tonne of coke and 0.20 per tonne of sinter); the EU ETS benchmarks, set on the best tenth of European plants, chain to about 1.4; and ArcelorMittal Dunkerque, at 6.8 Mt of steel and about 12 MtCO₂ a year, is at 1.76. The model charges the route's coal its own published factor, so this figure is never booked as such: steel_bf_process_residual keeps the remainder. The limitation: 1.76 and the 0.62 t of coal charged beside it describe a furnace about 20% better than the world average — worldsteel's route average is 780 kg of coal, which at this model's factor alone would release 2.08 t. The two are consistent with each other and should be changed together or not at all.

• IPCC 2006 Guidelines, Vol. 3, Ch. 4, Table 4.1 p. 4.25 — BOF 1.46, coke 0.56, sinter 0.20 tCO₂/t
• Commission Implementing Regulation (EU) 2021/447, Annex — hot metal 1.288, sintered ore 0.157, coke 0.217 allowances/t
• Direction générale des Entreprises (15 January 2024), France 2030 : soutien à la décarbonation d'ArcelorMittal Dunkerque — 6.8 Mt/y, ≈ 12 MtCO₂/y
• worldsteel, Raw materials — 780 kg of metallurgical coal per tonne of BF-BOF steel

What an electric furnace emits on site besides its electricity: the graphite electrodes it burns and the carbon charged to foam the slag and carburise the melt. The IPCC's Tier 1 default, and the floor of the published range — the EU ETS benchmark for carbon steel is 0.215, because it also covers the gas burners and the ladle furnace. It applies to both electric routes, from scrap and from hydrogen-reduced iron, since both end in the same furnace. Until 0.27.0 the model gave these two rows no direct term at all, so the only steel that emitted on site was the blast furnace's.

• IPCC 2006 Guidelines, Vol. 3, Ch. 4, Table 4.1 p. 4.25 — EAF 0.08 tCO₂/t
• Commission Implementing Regulation (EU) 2021/447, Annex — EAF carbon steel 0.215, high-alloy 0.268 allowances/t

Ethylene plus propylene produced in France in the base year: 2 270 kt of ethylene and 1 386 of propylene, from the SDES annual census of the chemical industry. It was 4 605 kt, and that figure was the wrong row of the right table. 2 270 + 1 386 + 949 = 4 605, where the 949 is the C4 cut, and the workbook took the line for the crackers' primary products as a whole. Three things in this model say the C4 does not belong in it: the row is named for ethylene and propylene, olefin_carbon_per_tonne is the 3.138 tCO₂/t of a CH₂ chain where butadiene is 3.255 and the cut also carries saturated butanes, and methanol_per_olefin is a methanol-to-olefins yield that makes light olefins and not a C4 stream. A perimeter that included the C4 would need the aromatics too, a different carbon coefficient and a different route — a different model, not a different number. The fleet has shrunk since, and this figure does not say so. France had six steam crackers; ExxonMobil shut Port-Jérôme-Gravenchon for good in June 2024, and output fell from 3 803 kt in 2021 to 2 878 in 2024 and 2 458 in 2025. Berre is under exclusive sale negotiation. The base year of this whole model is 2020, so a 2024 tonnage here would be a single recent observation among base-year ones; the fall is stated rather than substituted, and a 2050 scenario that assumes the 2020 fleet is making an assumption this why is where you find.

• SDES, enquête annuelle de recensement de la production industrielle — éthylène 2 270 kt, propylène 1 386 kt, coupes C4 949 kt (France, 2020)
• Citepa, Secten 2026, p. 6 — « la fermeture d'un vapocraqueur courant 2024 sur les six présents sur le territoire national »
• See docs/waste/olefins_and_plastic_demand.md for the full series 2014–2025 and the plant-by-plant fleet

The carbon a tonne of olefin can physically hold, which is the ceiling on any claim that the product stores CO₂. Ethylene and propylene are both 85.63% carbon by mass, so a tonne holds 0.8563 × 44.009 / 12.011 = 3.138 tonnes of CO₂ equivalent. Nothing about the route changes it: it is the polymer's own composition. It replaces a coefficient of 4.3 that was 37% above this ceiling. The old figure came from the teaching workbook's "Industry" sheet K48 with no derivation, and it reproduces exactly as the CO₂ fed to the front of the route rather than the carbon locked in the product: 2.988 tonnes of methanol per tonne of olefin (methanol_per_olefin, published) times the 1.3735 tCO₂ per tonne of methanol that stoichiometry demands is 4.105, and 3.138 / 4.105 = 0.765 is an ordinary light-olefin selectivity for a methanol-to-olefins step. The 0.97 tCO₂ of difference leaves as C₄⁺, coke and purge gas; it is consumed by the plant, not stored by the polymer, and crediting it was double-counting the losses as a removal.

• IPCC 2006 Guidelines, Vol. 3, Ch. 1 §1.3.1 — feedstock carbon is stored, and released when the product is oxidised
• Atomic masses: C 12.011, O 15.999, H 1.008 (IUPAC 2021)

France's 2020 cement production, to half a per cent: France Ciment publishes 16 422 kt. Inherited from the teaching workbook without a source and re-sourced in 0.27.0; the round figure is kept because the construction module is calibrated to it. Still the anchor, and no longer the driver. Since v0.20 cement volume is demand-driven and this constant sets no production directly; what it does is fix the total the four rows of construction_use must sum to, so that the base year of the industry chain, the process emissions and the kiln fuel all stay exactly where the workbook put them. It is a pre-2022 figure and the market has moved a long way since: France Ciment publishes 13.996 Mt produced and 15.970 Mt consumed in 2024, against about 19.1 Mt consumed in 2022. Re-anchoring it would re-anchor the whole industry block and is a change of its own.

• France Ciment / SFIC, Le ciment en chiffres 2022, p. 4 — cement 16 422 kt in 2020
• Teaching workbook, "Industry" sheet E54
• France Ciment, «Le ciment en chiffres 2025» — 13.996 Mt produced, 15.970 Mt consumed, 10.211 Mt clinker in 2024

The decarbonation of the limestone, and nothing else: 6 193 ktCO₂ of calcination in the French inventory for 11 759 kt of clinker in 2020. The neighbouring years give 0.524 (2019) and 0.531 (2021), and it is what stoichiometry predicts: the IPCC's 0.51 × 1.02 for kiln dust = 0.52 for a 65%-lime clinker, 0.53 at 67%. No change of kiln fuel can remove it — only capture, or less clinker. It was 0.7925, and that was a total wearing the wrong name. The old value was 10.2 MtCO₂ over 12.87 Mt of clinker. 10.2 Mt is not the calcination line for any year France has published — that line ran 6.19 to 6.81 over 2015–2021 — but it is inside the range of the whole cement industry's CO₂, calcination and kiln fuel. And 12.87 Mt was not an observation: it is 16 500 × 0.78, this model's own cement anchor times the top of its own clinkerRate slider. The kiln fuel is charged separately through the energy account, so it was emitted twice — and carbonCapture, which multiplies this term, was capturing fuel CO₂ under a name that said it could not. The limitation. 0.527 is a French fleet average for one year, and it moves with the lime content of the clinker and with how much of the lime comes from slag or ash instead of carbonate: the IPCC's range is 0.47 to 0.53. It is also domestic clinker; about a quarter of the clinker behind French cement is calcined abroad, outside this perimeter.

• Conseil national de l'industrie / DGE (May 2023), Feuille de route de décarbonation de la filière Ciment, Tableau 2-1 p. 7 — calcination 6 809 / 6 193 / 6 489 ktCO₂ in 2019 / 2020 / 2021 (SECTEN, GEREP)
• France Ciment / SFIC, Le ciment en chiffres 2022, p. 4 — clinker 13 005 / 11 759 / 12 222 kt in 2019 / 2020 / 2021
• IPCC 2006 Guidelines, Vol. 3, Ch. 2, Eq. 2.4 p. 2.12 — EF = 0.51 × 1.02 = 0.52 tCO₂/t clinker
• See docs/industry/steel_and_cement_base.md, §4

• Teaching workbook, "Industry" sheet, steam rows 49 and 51

• Teaching workbook, "Industry" sheet, direct-heat rows 50 and 52

Residual hydrogen use in the food industry, unaffected by any lever.

Surface times surfacic need overstates the stock's real heat consumption, so the workbook scales the whole 2020 account by this one coefficient to land on the observed 359.34 TWh. It is a single stock-wide calibration, not a per-segment fudge: every segment carries the same factor, so the shape of the stock is untouched and only its level is set by observation.

• Teaching workbook, "Building heating" sheet B53

The winter power drawn by electric space heating in 2020. It anchors the 2050 peak: the model computes a peak-coincident electric load for both years from the same expression and scales this figure by their ratio, so running the 2020 stock through the calculation returns 40 GW exactly. The workbook's own version did not -- it divided by the 2020 useful heat rather than the 2020 peak load, and returned 36.8 GW for 2020.

• Teaching workbook, "Building heating" sheet B55

• Citepa, SECTEN 2026 — latest consolidated inventory year

• Citepa, SECTEN 2026

• Citepa, SECTEN 2026

4% of the 1990 level, the current SNBC 3 industry reduction.

• SNBC 3 ministry pages

• Citepa, SECTEN 2026

47% of the 1990 level.

• SNBC 3 ministry pages

• Citepa, SECTEN 2026

45% of the 1990 level.

• SNBC 3 ministry pages

• Citepa, SECTEN 2026

4% of the 1990 level. Note what this implies: about 3 MtCO₂e for the whole energy branch in 2050, against roughly 600 TWh of electricity. That is around 5 gCO₂/kWh at the stack — far below the 40 gCO₂/kWh life-cycle factor the game applies, because the two count different things. This contrast is the point of the national reconciliation.

• SNBC 3 ministry pages

• Citepa, SECTEN 2026 — UTCATF

The official pathway weakens the natural sink, it does not strengthen it.

• SNBC 3 ministry pages, UTCATF

A transparent closure residual, not a published sector target: −66 MtCO₂e of total 2050 absorptions less the −23 MtCO₂e natural sink.

The published SNBC 3 gross national total for 2050, rounded.

• SNBC 3 ministry pages

What the model now represents of the industry sector, on the inventory's combustion-plus-process basis and with 2020 emission factors. The five value chains account for 37.0 — steel 15.2, cement 8.4, food 5.8, olefins 5.8, ammonia 1.7, electricity excluded as the inventory excludes it — and the seventeen other branches for a further 23.6, from 7.9 TWh of coal, 8.3 of fuel oil, 68.8 of gas and purchased steam, 19.4 of biomass and 2.5 MtCO₂ of process emissions. Steel and cement were recomputed in 0.27.0 from the model's own chain rows at 2020 volumes: 7 830 kt of blast-furnace steel and 3 770 of electric, and 11 759 kt of clinker at 0.527 tCO₂ of calcination. They were 18.8 and 13.1, the workbook's own figures, which carried the 2015 steel and the cement's kiln fuel twice. Cement at 8.4 is below the 9.53 the inventory books for 2020, and the gap is the kiln fuel the chain row undercounts — about 1.1 MtCO₂, named in industry_chain. Compared with the 61.6 MtCO₂e the inventory books, the model's perimeter is now slightly the smaller, by about as much as that kiln fuel. The two are not the same object anyway: the manufacturing survey the rest of industry is built from is a 2019 base, industry emissions have fallen since, and SECTEN's industry sector also contains construction and refining, which the survey does not. The difference is reported as a diagnostic and is not added to anything.

• Teaching workbook, "Industry" sheet rows 59–63, for food, olefins and ammonia
• Steel and cement: this model's own chain rows at the 2020 volumes — see docs/industry/steel_and_cement_base.md
• Feuille de route de décarbonation de la filière Ciment (May 2023), Tableau 2-1 p. 7 — cement 9 530 ktCO₂ in 2020
• INSEE, enquête EACEI 2019, for the other seventeen branches

RTE gives the fuel-side efficiency potential for industry as a whole and does not break it down by branch, so a single ceiling applies to every industrial post. The same source retains 10% as the readily achievable part; the model exposes the full 19.8% and lets the effort lever say how much of it is captured.

• RTE, Bilan prévisionnel — groupe de travail Consommation, after CEREN

42.8 MJ/kg, the standard lower heating value of jet A-1.

• Standard lower heating value of aviation kerosene

The anchor for today's ticket. Everything that is not fuel is derived from it through the fuel share of operating cost, so an error here moves the whole non-fuel block.

• EASA, European Aviation Environmental Report 2025, fuel price 2023

Combustion only — neither the upstream chain nor non-CO₂ effects.

• ICAO; corroborated by the US Energy Information Administration

2020, the base year of the workbook's service demand, to 2050. It is deliberately not the same anchor as the efficiency horizon: consumption per passenger-kilometre is anchored on the 2024 statistic, demand on the 2020 workbook value.

2024, the latest year of the traffic series, to 2050.

The raw French statistic — 6.95 Mt of kerosene for 237.4 Gpkm in 2024. It is higher than the figure the model uses because it also carries the freight in the holds and reflects actual load factors. Corrected for both, the same series gives about 19.3 g/pkm in 2023, which is the range the model's own aviation rows sit in. A real ticket therefore emits more than the per-passenger-kilometre figure below suggests.

• DGAC mémentos de statistiques and CPDP kerosene deliveries, compiled 1950–2024

Lower heating values, used to turn POMMES prices per tonne into prices per MWh.

• Standard lower heating values for hard coal, natural gas and hydrogen

• Standard lower heating value of methane

• Standard lower heating value of hydrogen

• POMMES-INDUSTRY, import_hourly.csv, France 2050

• POMMES-INDUSTRY, import_hourly.csv, France 2050

• POMMES-INDUSTRY, import_hourly.csv, France 2050

• POMMES-INDUSTRY, import_hourly.csv, France 2050

• POMMES-INDUSTRY, import_hourly.csv, France 2050

• SDES, Prix du gaz et de l'électricité au premier semestre 2025

• SDES, Prix du gaz et de l'électricité au premier semestre 2025

7.75 c€/kWh for bulk pellets.

• Propellet, energy price index, Q2 2025

• POMMES-INDUSTRY, conversion_factor.csv

• POMMES-INDUSTRY, conversion_factor.csv

• POMMES-INDUSTRY, conversion_factor.csv

Structural concrete dosages run roughly 250 to 400 kg a cubic metre, and 300 is the middle of the structural range. It is not the 266 kg a cubic metre that the cement-and-concrete literature uses for "béton équivalent": that figure is a whole-economy bookkeeping factor that already absorbs mortars, renders, screeds and bagged cement, and applying it to building cement alone overstates the concrete by about seven tenths. Used only to express the construction module's cement as a concrete tonnage in the materials account; no emission depends on it.

• Usual structural concrete dosage range; cross-checked against The Shift Project's 266 kg/m³ béton-équivalent factor, which is a different perimeter

Ordinary reinforced structural concrete. Presentation only, like the dosage above.

• Standard value for ordinary reinforced concrete, used in the comparative life-cycle assessment this module's timber coefficients come from

• POMMES-INDUSTRY, conversion_factor.csv

• POMMES-INDUSTRY, conversion_factor.csv

The cement kiln fuel appears in the cost model but NOT in the physical model, which gives cement only its grinding electricity. Cement combustion CO₂ is therefore missing from the emissions account — a known defect of the workbook, recorded here rather than silently patched.

• POMMES-INDUSTRY, conversion_factor.csv, solid_fuels mode 1

reference_plant_CCS less reference_plant.

• POMMES-INDUSTRY, conversion_factor.csv

• POMMES-INDUSTRY, conversion_factor.csv, steam methane reforming

• POMMES-INDUSTRY, conversion_factor.csv

• POMMES-INDUSTRY, emission_factor

• POMMES-INDUSTRY, conversion_factor.csv, methanol-to-olefins

France's total residential and tertiary floor area, kept as a cross-check rather than as an input: the model's own heated stock is 3 654.9 Mm², and building_surface_coverage reports the 87% ratio. The residential share that used to sit beside it is gone -- the stock carries the building type, so the split is counted rather than assumed.

• ADEME BatiZoom, after CEREN — residential and tertiary floor area

Demand reduction achieved by one deep renovation, used to convert the average stock improvement into an equivalent number of deep renovations.

• ADEME orders of magnitude for a BBC-rénovation-level retrofit

• Conventional lifetime for building-envelope work

• Conventional lifetime for a residential heat pump

Air-to-water heat pump, in the 60–100 €/m² range.

• ADEME cost ranges for residential heat pumps

Reduced VAT rate of 5.5% on renovation work.

• CSTB / OptoBat convention, VAT_RENOVATION = 1.055

• INSEE Focus 332 — main residences on 1 January 2024

Net purchase 1 459 + insurance 518 + maintenance 564.

• INSEE Première 1855 (2017), average household car budget

• INSEE Première 1855 (2017)

• SDES, Chiffres clés des transports 2026

The car fleet the model computes at the reference scenario, used as the denominator of the fleet ratio so household ownership cost scales with fleet size. It is pinned rather than recomputed because the model runs in one pass; the regression test checks it still matches.

New forest does not store carbon the year it is planted. Hectares planted less than ten years before the horizon are left out of the afforestation term altogether, which is a crude step where the truth is a curve; the alternative — a growth function nobody in the sources publishes — would be a curve we invented.

New forest is booked at the expansion rate rather than at the average per-hectare rate of the standing forest, because a young stand does not store like a mature one. The same source gives 5.0 for a renewal plan on existing forest and 0.2 to 2.4 for the rest of the forest depending on management and climate; 3.0 is the expansion line.

• IGN–FCBA (2024), Projections des disponibilités en bois et des stocks et flux de carbone du secteur forestier, table 15

1.0 tC/ha/y lost for twenty years when permanent grassland is ploughed, converted here at 44/12. The interval on it is ±40%, and the stock difference between the two uses (84.6 against 51.6 tC/ha over 0–30 cm) would imply 1.65 tC/ha/y if all of it were lost before a new equilibrium — it is not, and the modal value is what the inventory uses. These coefficients are measured on one country's soils and a port should check them against its own; they are shared because soil chemistry does not stop at a border, not because they are beyond argument.

• Citepa for the Haut Conseil pour le Climat (2021), Incertitudes liées à la comptabilisation des puits de carbone dans les sols, fig. 12 — after Arrouays et al. (2002)

0.5 tC/ha/y regained for twenty years when arable land goes back to grass, converted at 44/12. Half the loss rate, and deliberately so: the same source's finding is that "loss is twice as fast as gain", which is what makes re-grassing a slower repair than ploughing was a break.

• Citepa for the Haut Conseil pour le Climat (2021), Incertitudes liées à la comptabilisation des puits de carbone dans les sols, fig. 12 — after Arrouays et al. (2002)

How long a hectare goes on emitting, or storing, after it changes use. Twenty years is the tail the inventory applies, so at a horizon 26 years away only the last twenty years of conversions are still in the flux.

• Citepa for the Haut Conseil pour le Climat (2021), Incertitudes liées à la comptabilisation des puits de carbone dans les sols, tables 7 and 9

France carries the land module, so the natural sink is computed and the naturalSink slider is retired — declared, hidden, and inert. The three provisional editions carry 0 and keep the slider.

2024 to 2050. The stock the account starts from is Teruti's 2023 survey — the last edition before IGN's OCS GE takes over, and a definitional break to expect — and the flows run from the year after it. It is deliberately not the model's own 2020 base year: a land account is only as good as the survey it is written on, and this one is dated 2023.

87.9 ± 1.3 Mm³/y over 16.6 Mha. It was 5.8 m³/ha/y over 2005–2013: gross production is already falling, before any of the climate cases is applied.

• IGN, Mémento de l'inventaire forestier 2024, pp. 38–39

15.2 ± 0.6 Mm³/y over 16.6 Mha, windthrow excluded (4.2 Mm³/y more). It was 7.4 Mm³/y over 2005–2013: mortality has doubled in a decade, spruce 2.2, chestnut 1.6 and ash 1.4 Mm³/y of it. Half a percent of the standing stock a year, and the reason the flux balance has halved.

• IGN, Mémento de l'inventaire forestier 2024, pp. 34–35

The forest available for wood production, out of a forest area of 17.5 Mha. It is not the forest row of land_class, which is Teruti's 17.521 on Teruti's nomenclature: two perimeters, kept apart on purpose, and afforestation adds hectares to the land account without adding any to this one.

• IGN, Mémento de l'inventaire forestier 2024, p. 18

1 842 Mm³ of broadleaf and 985 of conifer, 2023. A thousand million cubic metres more than in 1985, 260 of them in the last ten years — which is the context for every flow above: the annual balance is under one percent of this stock, and a forest that stops growing does not shrink, it stops absorbing.

• IGN, Mémento de l'inventaire forestier 2024, pp. 22–25

Removals of live trees, ± 2.9, over 2014–2022 — 23.8 broadleaf and 29.4 conifer, up from 47.2 over 2005–2013. Not the 38 Mm³ of the commercial harvest statistic: the difference is firewood cut and never sold.

• IGN, Mémento de l'inventaire forestier 2024, pp. 36–37

IGN's own pair: 39 MtCO₂/y of net absorption by living biomass for a +19.5 Mm³/y flux balance. Both numbers come from the same document, the same campaigns and the same perimeter, which is what makes the ratio internally consistent for this identity. The research note behind this module recommends 1.5, and we depart from it. 1.5 is the SNBC 3's gross increment, "about 130 MtCO₂e", over IGN's gross production of 87.9 Mm³ — a gross ratio applied to a net balance. At 1.5 the identity gives 29.9 MtCO₂/y where IGN publishes 39, a 10 Mt hole that would then have to be absorbed by a calibrated constant. At 2.0 it gives 39.9. The marginal response of this pool, 2.0 tCO₂ per extra cubic metre harvested — 1.87 for the whole land account, once the wood-products pool that cubic metre would have fed is netted — sits between the 1.4 the IGN–FCBA B1→B2 pair implies and the 2.2 ADEME's S1–S4 spread implies, which is where a marginal coefficient should sit. This is the single most consequential number in the sink block and it is a live argument; 1.5 is recorded here so the argument stays visible. The "10 Mt hole" is a weaker argument than it looks, and it is recorded as such: the litter-and-soil residual already is a calibrated constant, and re-closed at 14.59 it holds the base-year forest line exactly at 1.5 too. What does separate the two is 2050. At 1.5, with the residual re-closed, the reference's natural sink is 32.3 MtCO₂e against the SNBC 3's own 23, because the residual does not weaken while the living biomass does. 2.0 stays until the forest line can be closed in level and in slope from the same pair; docs/forest/harvest_must_cost.md in the repository, stage B.

• IGN, Mémento de l'inventaire forestier 2024, pp. 40 and 45 — 39 MtCO₂/y for a +19.5 Mm³/y balance

Fitted so the dead-wood pool is 10.0 MtCO₂/y at the base year's mortality, which is what SECTEN's 2025 edition added to the forest line when it made the pool dynamic. Citepa does not publish the pool split in tonnes — that is in the ministry's CRT submission, not in SECTEN — so this level is calibrated against a difference rather than read off a table, and it is one of the two places the forest line rests on an inference.

• Citepa, Rapport SECTEN édition 2026 — forest line and its 2025 dead-wood addition

What closes the forest line: 39.88 of living biomass plus 10.00 of dead wood plus 10.0 of French Guiana leaves 4.62 against Citepa's −64.5 for 2024. It stands for litter and forest soil on land-use change, which is what the inventory books there, and its size is the measure of what the volume identity does not explain — 7% of the forest line. The ministry's CRT tables would replace it with a measurement.

• Citepa, SECTEN 2026 long series — UTCATF forests, −64.5 MtCO₂e in 2024

French Guiana's forest, 7.9 Mha of it, which the inventory treated as carbon-neutral until 2025 and now books at "about −10 MtCO₂e/y". It sits outside the metropolitan land account and inside the national total, so it is carried as a constant nothing steers — a fifth of the forest line that no lever in this game can touch, which is worth knowing before arguing about the other four fifths. No published figure with a decimal has been secured.

• Citepa, Rapport SECTEN édition 2026 — Guiana's undisturbed forest added to the inventory, order −10 MtCO₂e/y. No itemised figure published

Sawlogs in the 2024 commercialised harvest, round wood over bark.

• Agreste, Récolte de bois et production de sciages en 2024, CD 2025-17

Industrial wood — pulp and panels — in the 2024 commercialised harvest.

• Agreste, Récolte de bois et production de sciages en 2024, CD 2025-17

Energy wood sold, 2024 — 10.4 in 2023, the year it first exceeded industrial wood. It is well under half of what France actually burns.

• Agreste, Récolte de bois et production de sciages en 2024, CD 2025-17

53.1 of removals less 38.0 of commercialised harvest. Nothing measures it directly: the independent estimates run 14.5 Mm³ (IGN less the annual branch survey, 2016) to about 17 (ADEME's 16.9 Mm³ of self-consumed wood in 2018, three quarters of it from forests), and the last field study is from 2018. It is roughly 28% of the country's harvest and about 32 TWh of its wood energy, which makes it the largest energy flow in this model that nobody meters.

• IGN Mémento 2024 (removals) against Agreste CD 2025-17 (commercialised harvest) — the difference
• Solagro for ADEME, Étude sur le chauffage domestique au bois — approvisionnement (2021), p. 45

Zero. The French harvest statistic has no unutilised category: Agreste counts what leaves the forest gate, and the storm and dieback wood that is felled and abandoned is inside the mortality the forest inventory measures rather than inside the harvest. The four French uses of the harvest therefore close on the harvest without a fifth row, and forest_unutilised_share_base is exactly 0, which leaves the wood supply the number it was.

• Agreste, Récolte de bois et production de sciages en 2024, CD 2025-17 — the harvest is reported by destination, with no unutilised category

One. forest_harvest_base is already declared in the inventory's own bois fort tige — IGN's removals, not Agreste's round wood — so the harvest, the gross production and the mortality are three numbers in one volume convention and the identity needs no conversion. Stating it as a constant rather than leaving it implicit is what lets Germany, whose harvest statistic is published in Erntefestmeter ohne Rinde, keep its lever in the unit its own statistics use.

• IGN Mémento 2024 — removals, gross production and mortality are published in the same bois fort tige volume

Fitted to the SNBC 3's own statement: 22.5% to 30% of a 60 Mm³/y harvest should give at least 3 MtCO₂e/y in 2030. Measured against the base year's 11.95 Mm³ of long-lived product, that is 3.4 MtCO₂ over 6.05 Mm³ of extra volume, hence 0.562 tCO₂ per cubic metre a year. It is a flow coefficient standing in for a stock model with half-lives from seven to fifty years, and it is the crudest object in the sink block.

• Projet de SNBC 3, 31 October 2024, pp. 88–96 — wood-products sink ≥3 MtCO₂e/y in 2030

Sawn timber 9.5% plus panels 13% of the harvest, the SNBC 3 draft's own base-year split.

• Projet de SNBC 3, 31 October 2024, p. 89

Measured on one building and cross-checked against element coefficients, because no whole-building intensity by structural system is published anywhere. The anchor is an eight-storey, 142-dwelling mass-timber residential building whose bill of materials was compared with a functionally equivalent concrete one: 519 to 553 kg/m² of concrete avoided over 13 766 m² of gross floor area, which at 280–320 kg of cement a cubic metre and a density of 2 380 kg/m³ is 61 to 74 kg of cement, and 67 is the middle. A timber building is not a building without concrete, and that is the point of a saving rather than a substitution. The same measured building still carried 288 kg/m² of concrete in its foundations, its ground floor and its toppings, and used more lean concrete than its concrete twin. Against the 137 kg/m² this model books for new French housing, 67 is close to half — which is the right order: négaWatt reaches −46% of concrete at 80 to 95% timber, and half of 90% is 45. The competing way to model this is a displacement factor in tonnes of carbon avoided per tonne of carbon in the wood, and it is deliberately not used here. That literature has moved from 2.1 (Sathre and O'Connor 2010, 21 studies) to 1.2 (Leskinen 2018, 51 studies) to 0.55 at market level (Hurmekoski 2021, 44 studies, range 0.27–1.16), with a French critique arguing the benefit has been overstated several fold. A kilogramme of cement not made is a physical quantity this model already prices and emits; a displacement factor is an argument about counterfactuals. The model takes the physical route and leaves the argument to the annex.

• FPInnovations / Gouvernement du Québec, comparative LCA of an 8-storey mass-timber residential building (Arbora, Montréal) — bill of materials
• Hurmekoski et al. 2021, «Substitution impacts of wood use at the market level», Environmental Research Letters — market-level displacement factor 0.55 (0.27–1.16)
• FCBA/BIPE 2019 for CODIFAB, FBF and ADEME — element-level equivalences: 1 kg of wood replaces 5.24 kg of concrete and 0.18 kg of rebar in a load-bearing CLT façade

From the same measured building, and deliberately the net figure: the timber structure removes about 21 kg/m² of reinforcement, and adds back steel balconies, galvanised studs and brick support, so the whole-building balance is roughly −17 kg/m². Against the 21 kg/m² this model books for new French housing that is most of it — which is exactly why the national effect is small: new buildings are about a tenth of French steel, so even a wholesale switch to timber moves the steel bar by a few percent. ADEME's own biosourced scenario finds −2% then −5% on steel against −2% then −8% on cement, and this model reproduces that asymmetry rather than asserting it.

• FPInnovations / Gouvernement du Québec, comparative LCA of an 8-storey mass-timber residential building — whole-building steel balance
• ADEME 2019, «Prospective de consommation de matériaux pour la construction des bâtiments», scénario «développement des biosourcés»

Two independent anchors agree, which is the only reason this number is here at all. The measured mass-timber building carries 2 597 m³ of cross-laminated timber, glulam beams and glulam columns over 13 766 m² — 0.189 m³/m². The French biosourced-building label's top level asks for 45 kgC/m² for housing, which at the carbon content and density of softwood is 0.200 m³/m². Light timber frame sits well below both, so 0.189 is a mass-timber figure and overstates what an ossature-bois house uses; the model says so rather than splitting a coefficient it cannot source.

• FPInnovations / Gouvernement du Québec, Arbora mass-timber building — 2 047 m³ CLT + 378.2 m³ beams + 171.6 m³ columns over 13 766 m²
• Arrêté du 2 juillet 2024, label «bâtiment biosourcé» — 45 kgC/m² for level 3 in housing

A sawmill turns roughly half of a sawlog into sawn timber; the rest is slabwood, sawdust and bark, and most of it is sold as chips, pellets or panel furnish rather than lost. The factor is what lets construction timber demand be compared with the long-lived harvest the forest account already computes, and 2.0 is the round number the trade uses. It is provisional because a national yield is not published on the same perimeter as this model's harvest.

• Agreste, «Récolte de bois et production de sciages en 2024» — 18.3 Mm³ of sawlogs against 8.16 Mm³ of sawnwood produced

12% of new floor area framed in timber, and the weighting is the whole difficulty. The national timber-construction survey counts housing in dwellings and non-residential in square metres, and the two cannot be added: 6.6% of new dwellings in 2024 (18 250 of them) against 17.6% of new non-residential floor area (2.56 Mm²). Weighting the two observed shares by the floor areas this country file declares — 20.4 and 19.9 Mm²/y — gives 12.0%, and that is the number here. It treats the housing share by dwelling as if it were a share by area, which is an approximation the survey does not let anyone avoid. The detail is worth reading before quoting it: detached houses in the individually-commissioned sector are at 9.1%, collective housing at 5.6%, agricultural buildings at 26.1% and industrial ones at 19.9%. French timber construction is mostly sheds and barns, not apartment blocks. And the survey's own definition excludes roof trusses and external insulation: it counts buildings whose structure is wood, not the great majority of French houses that carry a timber roof on masonry walls.

• Enquête nationale de la construction bois, activité 2024 (8ᵉ édition), Xerfi Specific pour France Bois Forêt et le CODIFAB

A source of 0.4 MtCO₂e in 2024, and the sign is the news: this pool absorbed 4.7 MtCO₂e in 1990 and has drifted to zero and past it as short-lived products replaced long-lived ones. A country that keeps cutting and stops sawing turns its wood-products pool into an emitter, and France has.

• Citepa, SECTEN 2026 long series — UTCATF harvested wood products, +0.4 MtCO₂e in 2024

The national inventory report's 2021 inflow to the long-lived wood-products pool — sawn timber 1 158 536 tC, panels 1 360 387, plywood 209 296, 2 728 219 tC in all, 10.00 MtCO₂ — over this model's base-year long-lived volume, 53.1 Mm³ × 22.5% = 11.95 Mm³. That is 0.228 tC/m³, against the IPCC's default carbon density of sawnwood, 0.229 tC/m³ — agreement to half a per cent, unfitted, which is the cross-check that the strategy's "sawn plus panels" share and the inventory's tonnes are the same object.

• Citepa, Rapport national d'inventaire CCNUCC, édition 2023 (March 2023), §6.10 Produits bois, Tableau 245 — production of wood products from domestic removals, tC, 2021
• IPCC (2019), Refinement to the 2006 Guidelines, Vol. 4, Ch. 12, Table 12.1 — sawn wood 0.229 Mg C/m³, wood-based panels 0.269

The IPCC 2019 Tier 1 half-lives — sawn wood 35 years, panels 25, and 30 for plywood as the French inventory reads it — weighted by the inventory's 2021 inflows to the three categories, as one decay rate: k = 0.02401 a year, a half-life of 28.9 years. The French inventory itself uses a finer set — 50 years for framing, 30 for flooring, 15 for joinery, 10 for furniture, 25 for panels — and does not publish the weights between them, so the IPCC defaults are used and the inventory's own list is recorded here. Paper is left out: at a two-year half-life it is at equilibrium with its own inflow and adds nothing to a 2050 flux.

• IPCC (2019), Refinement to the 2006 Guidelines, Vol. 4, Ch. 12, Table 12.3 — Tier 1 default half-lives: paper 2, wood-based panels 25, sawn wood 35 years
• Citepa, Rapport national d'inventaire CCNUCC, édition 2023, Tableau 244 — durées de demi-vie des produits bois: panneaux 25, contreplaqués 30, papier 7, emballages 3, ameublement 10, menuiseries 15, charpente 50, parquets 30

The inventory report's 2021 outflows from the three long-lived categories — sawn timber 989 713 tC, panels 1 193 460, plywood 170 198, 8.63 MtCO₂ in all — over the decay rate of hwp_half_life, which is the stock a first-order pool must hold to release that much: 359 MtCO₂, 98 MtC. The stock this model derives from the base-year balance is a fifth larger, 433 MtCO₂, because the 2026 inventory vintage books the pool as a source of 0.4 where the 2023 report booked a sink of 1.4 for 2021, and the balance is what the model is held to. Shown beside it; not fitted to it.

• Citepa, Rapport national d'inventaire CCNUCC, édition 2023, Tableaux 242 and 246 — HWP balance −1 368 ktCO₂e in 2021; end-of-life outflows by category, tC

5.7 MtCO₂e absorbed in 2024 over Teruti's 9.139 Mha of permanent grassland. Calibrated on Teruti's grassland deliberately: the farm survey reports 10.527 Mha and the SNBC 3 a third figure again, 7 150 kha of "productive" grassland, and a coefficient derived on one area and applied to another would not close. The livestock block of stage B reads the farm-survey area instead, and carries the 1.388 Mha reconciliation explicitly.

• Citepa, SECTEN 2026 long series — UTCATF grassland, −5.7 MtCO₂e in 2024

11.7 MtCO₂e emitted in 2024 over Teruti's 17.265 Mha of arable land. It already contains the historic grassland conversions and the drained organic soils, which is why the module books a conversion flux only for the change the player makes and not for the base year over again. Cropland has been a source in every year the inventory covers, and a large one: +24 MtCO₂e in 1990.

• Citepa, SECTEN 2026 long series — UTCATF cropland, +11.7 MtCO₂e in 2024

5.0 MtCO₂e in 2024 over Teruti's 52 kha/y. It is a standing emission per unit of annual flow, not a one-off per hectare, because sealing and the biomass it removes are booked over a twenty-year tail. Per hectare it implies 15–110 tCO₂ once, depending entirely on which land is taken — 4 tC/ha from cropland, 11 from grassland, and 81 tC/ha of standing trees if the hectare was forest — and 96 tCO₂ per ha/y of flow sits inside that range, which is the check that the coefficient and the rate are the same pair. Choose the rate before the coefficient: on the cadastral rate of ~20 kha/y the same line would imply 250 tCO₂ per ha/y and nothing would close.

• Citepa, SECTEN 2026 long series — UTCATF artificial areas, +5.0 MtCO₂e in 2024
• Agreste, Primeur 2025-1 — Teruti artificialisation 52 kha/y, 1982–2022 mean

Wetlands, other land and dams, a source of 1.2 MtCO₂e in 2024 and between 0.5 and 2.1 in every year since 1990. Nothing in the game moves it and no source says what would.

• Citepa, SECTEN 2026 long series — UTCATF wetlands, other land and dams, +1.2 MtCO₂e in 2024

Five tonnes of CO₂ equivalent a hectare a year, mostly methane — what a rewetted peat soil still emits once the oxidation of the peat has stopped. Nothing in this edition reads it: the French organic-soil areas are zero. It is declared with the figure the European rewetting literature converges on so that the constant carries a number rather than a placeholder, and it would have to be re-sourced on French sites before any French result depended on it.

• Not a French source: the order of magnitude the European rewetting literature converges on. Nothing in this edition reads it, because the declared organic-soil areas are zero.

Zero, and not an approximation: the drained area France declares is zero, so the share of it already rewetted is zero too.

• countries/FR/FR.yaml, land_class — the declared drained organic-soil area is zero on every row, so this share has nothing to be a share of

Zero, for the same reason the areas are: Citepa's agricultural-soil N₂O already contains whatever the country's drained organic soils mineralise, inside the line ef_other_crop_n2o is calibrated on, and there is no published organic-soil area to take it back out with. With the area at zero this factor cannot move a French number whatever it were set to; it is declared 0 so that the two statements agree.

• Citepa, Secten 2026 — agricultural soil N₂O is published as one line, with no organic-soil split to take out of it

Cover crops extended 2.02, in-field agroforestry 1.10, temporary grassland in rotations 0.76 and hedges 0.15 — 4.03 MtC/y on arable land, converted at 44/12. Reduced tillage is excluded: INRAE reads its +0.68 MtC/y as a redistribution down the soil profile rather than a gain, and including it would add 2.5 MtCO₂/y here. The widely quoted "+21 MtCO₂/y, 4 per 1000" headline includes both no-till and forest land and is a different quantity.

• INRAE, Stocker du carbone dans les sols français ? (2019), résumé, table 2

Grassland intensification, 0.69 MtC/y on 3.9 Mha, converted at 44/12. The two potentials together are 17.31 MtCO₂/y, which is the study's own agricultural total without no-till. The module specification proposed splitting that total "arable 16.0 / grassland 3.8"; those two sum to 19.8, which is the potential with no-till, so the itemised split is used instead. The total is identical either way and only the pool the chart draws it in changes.

• INRAE, Stocker du carbone dans les sols français ? (2019), résumé, table 2

Roughly three quarters of the farmland France lost between 1982 and 2018 went to artificialisation and the rest to natural regrowth, so three quarters of what is artificialised is taken from arable land and the rest from heath and scrub. It decides which class shrinks, and therefore how much of the cropland source the artificialisation lever removes as it goes.

• Agreste, L'occupation du sol entre 1982 et 2018, Dossiers 2021-3

Zero. The ZAN accounting France's land-take figures come from does not separate grassland from arable land in what is built on — both are espaces naturels, agricoles et forestiers — so the arable share carries the whole farmland side and this term is exactly zero, leaving the account as stage A wrote it. What would close it: Teruti's own year-on-year transition matrix, which does distinguish the two.

• Agreste Teruti 2023 — land take is measured against espaces naturels, agricoles et forestiers as one aggregate

Zero, and for the same reason: the quarter of French land take that is not booked to arable land is booked to the semi-natural class, which is where Teruti puts the woodland edges, the scrub and the bare ground that roads and estates actually take. Splitting a forest share out of it would move hectares between two classes of the same account without a published measurement to do it with.

• Agreste Teruti 2023 — the semi-natural class is where the survey puts the woodland edges, scrub and bare ground land take actually consumes

Teruti's 1982–2022 mean, and the rate artificialisation_carbon_content is derived on — the two have to be the same pair or the artificial pool does not close. The recent rate is under 38 kha/y on the same measure, and 15–23 on the cadastre, which counts different hectares.

• Agreste, Sols artificialisés : +66 % en 40 ans, Primeur 2025-1

Deliberate afforestation outside existing forest was 100 ha/y in 2021 — 0.1 kha/y, which rounds to nothing at the precision this account is kept in. It is not the forest expansion the inventory measures, 90 kha/y, which is mostly spontaneous and already inside the standing forest's own sink.

• Projet de SNBC 3, 31 October 2024, p. 92 — afforestation outside forest, 100 ha/y in 2021

Zero because the SNBC 3 holds the permanent grassland area, not because the countryside is stable: France ploughed about 64 kha/y through the 2000s. This is the strategy's position taken as the base, and it is one of the places where the reference scenario on this tab is already an effort.

• SNBC 3 — permanent grassland maintained to 2050

Zero by construction: INRAE states its potential as storage additional to current practice, so the base year has taken none of it by definition.

• INRAE, Stocker du carbone dans les sols français ? (2019), résumé

Living biomass, dead wood, litter and soil on land-use change, and French Guiana, in one line. It was −80.9 in 2008. Forest figures are five-year centred means extrapolated for the last two years, so 2023 and 2024 will be revised again.

• Citepa, SECTEN 2026 long series, sheet BDD_Secten, UTCATF forests

A source in 2024, against −4.7 in 1990.

• Citepa, SECTEN 2026 long series, sheet BDD_Secten, UTCATF harvested wood products

The second of the two lines that absorb, and it was −8.6 in 1990.

• Citepa, SECTEN 2026 long series, sheet BDD_Secten, UTCATF grassland

A source, and the largest one in the land account. +24.0 in 1990.

• Citepa, SECTEN 2026 long series, sheet BDD_Secten, UTCATF cropland

A source, and remarkably steady: between +4.2 and +5.5 every year since 1990.

• Citepa, SECTEN 2026 long series, sheet BDD_Secten, UTCATF artificial areas

The six lines sum to −51.90 while Citepa's own UTCATF total for 2024 is −51.96. The 0.06 is the rounding of the published sub-sector lines against their published total, and land_sink_check_2024 reports it rather than absorbing it into a coefficient.

• Citepa, SECTEN 2026 long series, sheet BDD_Secten, UTCATF wetlands, other land and dams

One, by definition: dietDairy is an index on the base year, so the base year sits at 100%. It is declared rather than written into the formula so that the switched-off branch of the module names what it reads.

Zero: the base year has cut none of its own waste, because the waste share is measured on it.

One: the export volumes the table declares are the base year, so the index that scales them is one there.

One, like livestock_export_base: the export hectares the crop block declares are the base year's, and the index that scales them is one there.

One: mineral_n_base is the base-year delivery, and nIntensity is a percentage of it.

Zero, and this is a statement about the calibration rather than about farming practice: the per-head emission factors are fitted to the observed inventory year, so whatever low-methane feeding that year already contained is inside the factors. The lever measures the change from there, and the change at the base year is nothing.

Zero, for the same reason as enteric_mitigation_base: the digesters the base year already runs are inside the observed emission factors, and the lever books the change from there. It is not a claim that no manure is methanised today.

Zero: the base year burns all of the fossil fuel the inventory measures on its farms.

The observed emission factor of fossil liquid fuel, the value efLiquid slides away from and the one the base-year farm burned. It is used in one place only — to say how much energy the base year's farm fuel represents — and no emission is charged at it.

The natural logarithm of two, which turns a half-life into a first-order decay rate: k = ln 2 / half-life. Declared once rather than written into a formula, like the nitrogen fraction of ammonia, because a number that appears in an equation should be a number a reader can find.

• IPCC (2006), Guidelines for National Greenhouse Gas Inventories, Vol. 4, Ch. 12, Equation 12.1 — k = ln(2)/HL

14 ÷ 17: the nitrogen in a tonne of ammonia, from the atomic masses. It is the one number in this module that is chemistry rather than statistics, and it is what turns a nitrogen demand into an ammonia tonnage for the industry chain.

• IUPAC atomic weights: N 14.007, H 1.008 — 14.007 / 17.031 = 0.8225

5 827 kt of meat consumed in 2024 at 85.0 kgec a head gives 68.55 million people. It is derived from the food balance itself rather than taken from the census, because a diet in kilograms per person and a population from another source will not multiply back to the published tonnage — and the tonnage is what sizes the herd.

• Agreste/FranceAgriMer, Synthèses n°443 — 5 827 kt of meat, 85.0 kgec per head in 2024

69 206 324 people at 1 January 2050 in the central scenario of INSEE's 2021–2070 projections, France including Mayotte, from the detailed results workbook. The projection peaks in 2044 at 69.3 million and is back to 68.1 by 2070 — the figure this constant carried as a provisional value until stage E, mislabelled as 2050. It scales every diet-driven quantity by about one per cent over the base year's 68.55, which is itself derived from the food balance rather than from the census; the projection's own 2024 figure is 67.8 million, so on the projection's own base the increase would be two per cent, and the difference is the perimeter of the balance sheet, not the demography.

• INSEE, Projections de population 2021–2070 pour la France, scénario central — résultats détaillés (00_central.xlsx), sheet 'population', Total, 1 January 2050: 69 206 324
• INSEE Première n° 1881 (November 2021), Projections de population 2021-2070 — 69.3 million in 2044, 68.1 million in 2070

Beef and veal 20.8 + pork 30.6 + sheep 2.1 kgec a head in 2024, on a total meat consumption of 85.0 kgec. Red meat has fallen 5.8% a head over twenty years while its share of all meat went from three quarters to two thirds.

• Agreste/FranceAgriMer, Synthèses n°443 — consommation de viande 2024, table 1

2 133 kt of poultry consumed in 2024 over 68.55 million people. The balance sheet publishes the tonnage and the +7.1% on 2023 but not the per-capita figure, so this one is the division.

• Agreste/FranceAgriMer, Synthèses n°443 — 2 133 kt of poultry (rabbit included) in 2024

3.8 Mt of edible food waste — 55 kg a head — on a food supply of order 55 Mt. Read it against the two larger figures it is constantly confused with: total food waste on the European definition is 9.7 Mt, and ADEME's often-quoted "30%" is the share of the chain's losses that occur at consumption, not a share of the food. Using either of those here would make the waste lever look several times as powerful as it is; the finding that food waste is a small lever on this account depends entirely on this 7%, and the SDES detail against a French food balance sheet is what would confirm it.

• SDES, Les déchets alimentaires en France et dans l'UE en 2023 — 9.7 Mt, of which 3.8 Mt edible (55 kg/capita)

The number the whole milk–beef coupling rests on, and it is an estimate. The farm survey reports 501 kt of cull-cow carcass in the 1 267 kt of French beef, and this module reads that as the dairy herd's contribution — 40%. The survey's line is all cull cows, dairy and suckler together, so the reading is generous to the dairy herd; Idele's herd-flow accounts would settle it. The alternative split, proportional to cow numbers, gives yields of 74 and 282 kg a head instead of 165 and 207, and a dairy-only diet cut would then send far less beef to market.

• Agreste, Statistique agricole annuelle 2024 définitive, pp. 23–25 — 1 267 kt of beef and veal, of which 501 kt cull cows

−14% of enteric methane where a lipid-enriched ration is fed, the figure the SNBC 3 uses after Pellerin's 2013 assessment for INRA. On 82% of housed cattle it is the −5% herd-wide the strategy books. 3-NOP would give 20–35% but is not in the strategy and is fed only during the housed period, so it is named in the lever's why rather than used here.

• SNBC 3 (July 2026), Partie 1, p. 128, citing Pellerin et al. (INRA 2013)

Three fifths of the methane a manure store would have released is captured when that manure goes to a digester instead. No French figure was secured for this, and it is one half of a provisional pair — the other is manure_ch4_share in the livestock table. Together they are why manureMethanised defaults to zero rather than to the SNBC 3's 80%: booking three megatonnes of abatement on an unpublished split would have closed most of this module's gap to the strategy with a coefficient nobody can check. Citepa's OMINEA database, or the CRF table 3.B for France, would replace both.

• NO PRIMARY SOURCE SECURED. The order is the one the methanisation literature quotes for covered storage plus digestion; Citepa OMINEA would replace it

HFC leakage on farms, the one line of the "other livestock" block that is not a ruminant. Citepa books it inside agriculture, so the module carries it there; it is subtracted from the small-ruminant factor's calibration so it is not counted twice.

• Citepa, SECTEN 2026, Agriculture-Emissions — other livestock 4.36 MtCO₂e of which HFC 0.02

1 817 kt of nitrogen delivered in the 2024 calendar year — ammonium nitrate 631, urea 520, UAN 481, compounds 120, other 65 — which is the series the inventory's own emission factors are implied against. The fertiliser survey publishes a July–June campaign instead: 1 887 kt for 2024-25, up 2.8%, and 1 835 revised for 2023-24. The two are not the same quantity and mixing them would move every nitrogen result by about four per cent.

• Citepa, SECTEN 2026, Agriculture data file, sheet Compléments, rows 34–39 — calendar-year mineral N deliveries
• Agreste, Infos rapides Engrais, January 2026 — campaign 2024-25 at 1 887 kt N

735 kt of usable nitrogen collected in buildings and spread, out of 1 462 kt excreted in all — cattle 1 228, pigs 78, poultry 64, sheep 75. 92.8% of the 105 Mt of manure and treatment co-products is simply stored, which is the pool manureMethanised acts on.

• RMT MAELE, fiche 9 valorisation des effluents d'élevage (2022 data), table 7

727 kt of nitrogen deposited directly at pasture, the other half of the 1 462 kt excreted. It carries its own emission factor, a third above the one for spread manure, so the two are charged separately.

• RMT MAELE, fiche 9 valorisation des effluents d'élevage (2022 data), table 7

364 580 tonnes of nitrogen fixed biologically in 2023, the latest year of Eurostat's gross nitrogen balance for France, read through the API. The series is volatile — 405 kt in 2021, 290 in 2022, 365 in 2023, a 2019–2023 mean of 329 — because the legume harvest is, and the latest year is used rather than a mean so that the number is the one a reader can find in the table. It replaces the 300 kt this constant carried as an order of magnitude until stage E; the base-year crops block moves by 0.15 MtCO₂e with it, inside its tolerance, and the same table gives the mineral fertiliser at 1 733 kt for 2023 against the inventory's 1 817 for 2024, which is the calendar-year series this module reads.

• Eurostat, Gross nutrient balance (aei_pr_gnb), France, nitrogen, indicator I_OTH_BIO 'biological fixation of nitrogen' — 364 580 t in 2023 (updated 10 February 2026)

Sixty per cent more biological fixation when the legume area rises by the whole 1.7 Mha the INRAE credit is measured over — 180 kt of nitrogen, or about 106 kg a hectare, which is inside the range the fixation literature reports for faba bean and clover and below what a pure legume ley fixes. 1.0 to 2.7 Mha nearly triples the cropped legume area while the grassland legumes do not move, so the total rises by less than the crop area does. It pulls the other way from the mineral credit, and by more: 180 kt fixed against 128 kt of mineral nitrogen replaced, so the nitrogen balance grows as the fertiliser bill shrinks. Emissions still fall, because mineral nitrogen is charged at 5.44 tCO₂e a tonne and total input at 2.31, but the two move in opposite directions and the model reports both.

• NO PRIMARY SOURCE. Derived as an assumption from the legume areas of the SNBC 3 and the %Ndfa ranges of the fixation literature

One million hectares of legumes in the French rotation, the figure the SNBC 3 starts its 1 → 2 → 2.7 Mha trajectory from. The farm survey's protein crops and pulses are 321 kha of that; the rest is forage legumes.

• SNBC 3 (July 2026), Partie 1, p. 128 — legumes 1 → 2 → 2.7 Mha

128 kt of mineral nitrogen no longer needed, which is what INRAE books for the legume part of its −944 kt proposal. It is one of four terms there — organic 330, agroecology 426, legumes 128, volatilisation 60 — and the only one this module reads separately, because the other three are inside the nIntensity lever rather than beside it.

• INRAE, hypothèses AMS SNBC 3 (April 2023) — 128 kt N credited to legumes, on +1.7 Mha

The 1.0 → 2.7 Mha increase the 128 kt credit is booked over. Declaring the span beside the credit is what makes the pair a coefficient instead of two numbers; the module reads it linearly inside the span, which is what the study does, and the lever's bounds keep the scenario close to it.

• SNBC 3 (July 2026), Partie 1, p. 128 and INRAE AMS (April 2023) — legumes 1 → 2.7 Mha

5.6% of the field-crop area in organic farming in 2024, the SNBC 3's own base figure for the target this lever defaults to; Agreste's arable-land survey had 4.6% in 2022. Not the 10% of the agricultural area the organic agency publishes, which counts 2.7 Mha of mostly grassland.

• SNBC 3 (July 2026), Partie 1, p. 127 — organic farming 5.6% of the field-crop area in 2024
• Agreste, Primeur 2023-8, Enquête Terres labourables 2022 — 4.6% of field-crop area organic, about 600 000 ha

Two thirds of the conventional yield, at the scale of a rotation: the INRAE hypotheses for the strategy take organic yields at 60% of conventional now and 70% at the horizon, and 0.65 is the middle of that. It is the contested number in this block, and the range is wide. Agreste's field-crop survey measures the French gap crop by crop — soft wheat −57%, winter barley −47%, triticale −37%, maize −31 to −35%, sunflower −28% in 2022, stable over 2018–2022 — which at France's crop mix is worse than 0.65; the two global meta-analyses are better: Seufert et al. 2012 find organic yields 25% lower on average, from 5% to 34% depending on crop and practice, and Ponisio et al. 2015, on a data set three times larger, 19.2% ± 3.7% lower, and 8–9% where rotations and multi-cropping are used. A reader who takes Ponisio should read the land cost of the organic lever as about half of what this page shows; one who takes Agreste's wheat should read it as half as much again.

• INRAE, hypothèses AMS SNBC 3 (April 2023), p. 11 — 'rendements en bio : 60% puis 70% du conventionnel (à l'échelle de la rotation)'
• Agreste, Primeur 2023-8 (June 2023), Enquête Terres labourables 2022 — organic yield gaps: soft wheat −57%, winter barley −47%, triticale −37%, sunflower −28%; stable 2018–2022
• Seufert, Ramankutty & Foley (2012), Comparing the yields of organic and conventional agriculture, Nature 485, 229–232 — 25% lower on average, 5% to 34% by context
• Ponisio et al. (2015), Diversification practices reduce organic to conventional yield gap, Proc. R. Soc. B 282, 20141396 — 19.2% ± 3.7% lower; 8–9% with rotations and multi-cropping

A stance, bracketed by two published ones. A conventional hectare keeps its yield down to 90% of the 2024 mineral dose; below that, every cut is nitrogen the crop was using. INRAE's hypotheses for the strategy book 426 kt of their −944 kt as "mesures agroécologiques et optimisation fertilisation N" — 20% of the 2020 mineral dose — "sans altérer les rendements (ou marginalement)", with levers taken from INRA 2013, whose specification was no loss or less than 5%: realistic yield targets alone are 10 to 15% of the dose. That reading is a plateau at 0.80. The GRAFS curve with practices unchanged has no plateau at all, and the JRC's DayCent run finds soft wheat losing up to 2.1% of its yield for the first 5% of mineral nitrogen cut: that reading is 1.00. Mineral nitrogen has already fallen 13% since 2010, so part of the excess INRA measured in 2006 practices is gone; 0.90 sits between the two. At the reference dose of 70% the conventional hectares keep 0.90 of their yield; the plateau at 0.80 would leave them 0.94, and no plateau 0.87. The SNBC 3 itself books the whole cut at no yield cost, which is the one assumption this constant declines to copy.

• INRAE, hypothèses AMS SNBC 3 (April 2023), p. 11 — −944 kt N, of which −426 'mesures agroécologiques et optimisation fertilisation N'; pp. 38, 42 — levers 'sans altérer les rendements (ou marginalement)', after INRA 2013
• Pellerin et al. (2013), Quelle contribution de l'agriculture française à la réduction des émissions de GES ?, INRA, Action 1, p. 76 — yield targets overestimated by 10–20%; a realistic target saves 15–20 kg N/ha, 10–15% of the dose
• Pacifico et al. (2024), Quantifying the impact of an abrupt reduction in mineral nitrogen fertilization on crop yield in the European Union, Sci. Total Environ. 954:176692 — mineral N −5/−15/−25% gives per-crop EU yield losses up to 2.1/6.4/11.2%, largest for soft wheat (abstract)
• docs/agriculture/nitrogen_yield_response.md §4.1, §5 and §6.3 — the bracket, and the table of y(d) under each reading

Two thirds of the nitrogen French cropland receives leaves the field in the harvest. Lassaletta et al. (2014) put France at about 70% in 2009, on a path that rose from 30–35% in the 1960s — from the mid-1970s French yields went up on a flat or falling input — read on the first author's 2018 reproduction of their Fig. 1(c) rather than on the article itself. A cropland budget built from Eurostat's gross nitrogen balance for 2019–2023 gives 0.68, between 0.61 and 0.75 by how the inputs are split between cropland and grassland; the Seine basin gives 0.63 for 2014–2019. 0.67 is the rounded middle. It sets the curve's bend below the plateau: across 0.60 to 0.72 the conventional yield at the reference dose moves between 0.91 and 0.89.

• Lassaletta et al. (2014), 50 year trends in nitrogen use efficiency of world cropping systems, Environ. Res. Lett. 9:105011 — eq. 1, Y = Ymax·F/(F + Ymax); France NUE ≈ 70% in 2009 (Fig. 1(c), read on Lassaletta et al., INMS workshop, Edinburgh 2018, slides 5–6)
• Billen & Garnier (2023), The water-agro-food system: upscaling from the Seine river basin to the global scale, C. R. Géoscience 355:301–315 — eq. 1 p. 305; Fig. 2 p. 306, Seine cropland 2014–2019: harvest 496 kt N, losses 295 kt N
• Eurostat, gross nutrient balance aei_pr_gnb, France, nitrogen, 2019–2023 — inputs and outputs the cropland budget is derived from
• docs/agriculture/nitrogen_yield_response.md §3.3 — the derived cropland budget and its range

The soft-wheat chain of ADEME's 2016 loss study, compounded downstream of the farm: 6% lost in milling, 10% in baking and 8% of what is bought, so 1 − 0.94 × 0.90 × 0.92 = 22% of the flour equivalent that leaves the farm. Wheat stands for the plant-food basket because it is the largest item of it and the only one the study follows from grain to plate; potatoes lose more, sugar and oil less. The 6% lost in the field is not in it — it is inside the yields.

• ADEME (2016), Pertes et gaspillages alimentaires : l'état des lieux et leur gestion par étapes de la chaîne alimentaire — synthèse de la filière blé tendre–farine–pain: 6% / 6% / 10% / 8% des achats

The four shares are one arithmetic on two sources — the farm survey's 2024 areas and FranceAgriMer's five-campaign cereal balance — over the 16.33 Mha of arable land that is not growing fuel (16.948 less 0.618). Cereals, 8.527 Mha, are split by the balance: 44.2% exported (26.9 of 60.9 Mt), 29.1% fed (9.2 Mt to the feed industry and the 14% of the harvest consumed or stocked on farms), 22.0% milled, malted, starched, distilled or ground (4.6 + 1.6 + 4.3 + 2.1 + 0.6 + 0.2 Mt), 4.7% seed and stock change. Food is the milled share less the 0.191 Mha of ethanol cereals, plus oilseeds less the 0.4 Mha of biodiesel, plus beet less its ethanol, plus potatoes and vegetables: 4.40 Mha, 27.0%.

• Agreste, Statistique agricole annuelle 2024 définitive, pp. 10–19 — cereals 8 527 kha, oilseeds 2 275, protein crops 321, sugar beet 412, potatoes 218, vegetables 240, silage maize 1 275, temporary grassland 2 534, fallow 521, arable 16 948
• FranceAgriMer, Fiche filière céréales 2026 — averages 2020/21–2024/25: 60.9 Mt produced, 86% collected, 26.9 Mt exported, 9.2 Mt to compound feed, 4.6 Mt milled, 1.6 malted, 4.3 starch, 2.1 alcohol, 0.6 semolina, 0.2 maize milling

Silage maize 1.275 Mha and temporary grassland 2.534, plus 29.1% of the cereal area and the 0.321 Mha of protein crops: 6.61 Mha, 40.5% of the non-energy arable area — the largest use of French fields. Same two sources as arable_share_food.

• Agreste, Statistique agricole annuelle 2024 définitive, pp. 10–19
• FranceAgriMer, Fiche filière céréales 2026 — 9.2 Mt to compound feed, 86% of the harvest collected

44.2% of the cereal area, 3.77 Mha, 23.1% of the non-energy arable area: the hectares behind the 26.9 Mt France ships abroad in an average year. Oilseed exports are not in it — rapeseed is mostly crushed at home.

• FranceAgriMer, Fiche filière céréales 2026 — 26.9 Mt exported of 60.9 Mt produced

Fallow 0.521 Mha, 4.7% of the cereal area for seed and stock changes, and the 0.625 Mha the farm survey's arable total holds beyond the crops itemised here. It closes the four shares to one, and the rounding of the other three sits in it.

• Agreste, Statistique agricole annuelle 2024 définitive, pp. 10–19 — fallow 521 kha; arable 16 948

Silage maize and temporary grassland, 3.81 Mha, over the 6.61 Mha of feed area. Forage is eaten by cattle and follows the cattle index; the remaining 42% is grain and follows the compound-feed species mix.

• Agreste, Statistique agricole annuelle 2024 définitive — silage maize 1 275 kha, temporary and artificial grassland 2 534

Zero: no French hectare grows a main crop for a digester. The 2016 decree caps main-crop feedstock at 15% of a plant's tonnage and the sector's own feedstock surveys report the rest as manure, waste and cover crops, which is civeArea. The base-year biogas balance is therefore built without a main-crop term, and biogas_other carries whatever main-crop maize does reach a French digester inside the residual it already names.

• Décret n° 2016-929 du 7 juillet 2016 — main crops capped at 15 % of a plant's gross feedstock tonnage

Zero, because the area is. A yield beside an area of zero is a number nothing multiplies; it is declared 0 rather than at some agronomic value so that no future reader mistakes it for a French measurement.

• countries/FR/FR.yaml, energy_maize_area_base — zero, so this yield multiplies nothing

Zero. Citepa's agriculture sector has no line for the digestion of energy crops — the emissions of a French digester's store are inside the manure management the livestock block already carries — so there is nothing to book here and nothing to divide by an area that is itself zero.

• Citepa, Secten 2026 — the agriculture sector has no energy-crop digestion line; a digester's store is inside manure management

2 044 kt of poultry feed of 4 801 kt of compound feed in the last quarter of 2024, Agreste after the feed industry's own returns: poultry 42.6%, cattle 27.2%, pigs 22.5%, the rest 7.7%. A quarter rather than the year because the quarterly note is what the statistician publishes with a species split.

• Agreste, Infos rapides Alimentation animale, February 2025 — production of compound feed, October–December 2024: total 4 801 kt, cattle 1 304, pigs 1 081, poultry 2 044

1 304 of 4 801 kt, same note.

• Agreste, Infos rapides Alimentation animale, February 2025 — cattle feed 1 304 kt of 4 801 in Q4 2024

1 081 of 4 801 kt, same note.

• Agreste, Infos rapides Alimentation animale, February 2025 — pig feed 1 081 kt of 4 801 in Q4 2024

7.65 MtCO₂e of N₂O on the mineral-fertiliser line divided by 1 817 kt of nitrogen delivered, both from the same inventory year. It implies an emission factor of 1.0% of the nitrogen applied lost as N₂O-N, and the IPCC 2019 default EF₁ is 0.010, which gives 4.16 tCO₂e/t N — one per cent apart, and neither was fitted to the other. That agreement is the best cross-check in this module, and it is worth stating because the same inventory calls soil N₂O its most uncertain line.

• Citepa, SECTEN 2026, Agriculture-Emissions — mineral fertilisers 9.90 MtCO₂e in 2024, of which N₂O 7.65 and CO₂ 2.24
• IPCC 2019 Refinement, Vol. 4 Ch. 11, Table 11.1 — EF₁ = 0.010 (0.002–0.018)

2.24 MtCO₂ of urea hydrolysis and liming on the same line, over the same 1 817 kt of nitrogen. Liming is driven by area and soil pH, not by nitrogen, so charging it per tonne of nitrogen is a stated approximation: a scenario that halves the nitrogen dose here also halves the liming, which is not what a farm would do. The inventory publishes the two on one line, which is why they are carried on one factor.

• Citepa, SECTEN 2026, Agriculture-Emissions — mineral fertilisers and amendments, CO₂ 2.24 MtCO₂ in 2024

Zero, because France charges its liming on nitrogen instead: the 1.2328 tCO₂e per tonne of mineral N in ef_mineral_co2 is urea hydrolysis and liming together, as that constant's own why says. The two are alternatives — declaring a fixed liming line here as well would book the same limestone twice — and the choice is recorded rather than silent, because charging liming on nitrogen is an approximation and a reader should be able to find where it was made.

• countries/FR/FR.yaml, ef_mineral_co2 — the French liming CO₂ is inside that factor, so this line is zero by construction

1.48 MtCO₂e over 735 kt of nitrogen in spread manure — an implied emission factor of 0.48% of the nitrogen lost as N₂O-N, half the mineral one. The inventory applies a lower factor to organic nitrogen because it is released more slowly; the IPCC's own 2019 refinement makes the same distinction in wet climates.

• Citepa, SECTEN 2026, Agriculture-Emissions — organic fertilisers 1.48 MtCO₂e in 2024

1.45 MtCO₂e over 727 kt of nitrogen deposited at pasture. The IPCC's EF₃PRP for cattle is 0.004 in the default case — 1.67 tCO₂e/t N — and 0.006 in a wet climate — 2.50. France sits between the two, which is what a country with both an Atlantic west and a Mediterranean south should.

• Citepa, SECTEN 2026, Agriculture-Emissions — grazing deposits 1.45 MtCO₂e in 2024
• IPCC 2019 Refinement, Vol. 4 Ch. 11, Table 11.1 — EF₃PRP cattle 0.004, wet 0.006

The module's largest single approximation. 8.25 MtCO₂e of "other crop emissions" — residues, mineralisation, leaching and the indirect N₂O pathways — divided by the whole 3 643.6 kt of nitrogen input, so that the base year closes. Re-calibrated in stage E, from 2.305 on 3 579 kt, when the biological fixation went from a 300 kt order of magnitude to Eurostat's 364.6 — a calibrated factor follows the input it was fitted on, or the base-year check it exists for stops holding. It lumps an area-driven quantity (crop residues) with a nitrogen-driven one (indirect N₂O), and a scenario that cuts nitrogen hard therefore also cuts the residue term, which is not physical. The inventory's CRF table 3.D splits them, and reading it is what would replace this factor with two.

• Citepa, SECTEN 2026, Agriculture-Emissions — other crop emissions 8.25 MtCO₂e in 2024

Field burning of crop residues, down from 0.10 MtCO₂e in 1990 as the practice was restricted. It is a constant because no lever here drives it and because it is a fiftieth of the crops block.

• Citepa, SECTEN 2026, Agriculture-Emissions — residue burning 0.02 MtCO₂e in 2024

Engines, motors and boilers: agriculture 10.33 + forestry 0.41 MtCO₂e, of which 9.42 is CO₂. It is a seventh of the whole sector and it behaves like any other combustion — which is exactly why the SNBC 3 takes it to zero and why agriFuelSwitch defaults to 100%.

• Citepa, SECTEN 2026, Agriculture-Emissions — engines and boilers, agriculture 10.33 and forestry 0.41 MtCO₂e in 2024

10.527 Mha of permanent grassland in the farm survey less 9.138540 Mha of it in the land survey. The farm survey counts farm-declared rough grazing as permanent grassland; the land survey books the same hectares under heath and scrub. Both are right about their own nomenclature, and the module keeps them apart on purpose: the livestock block reads grassland + grassland_rough, so its base year is the farm survey's 10.527, and the land account still closes on the land survey's 54.919 territory. Merging them would break either the account or the calibration. It is also the largest of the module's three statistical reconciliations, and the definitional spread behind it is wider still — the strategy's draft says 9.6 Mha and its final text 7.15 Mha of "productive" grassland.

• Agreste, Statistique agricole annuelle 2024 définitive — permanent grassland 8 836 kha on farms + 1 680 kha outside, 10.527 Mha in all
• Agreste, L'occupation du territoire en 2023, Chiffres et données n° 10 — permanent grassland 9.138540 Mha

The ammonia French chemistry makes for something other than fertiliser — nitric acid for explosives, caprolactam, industrial refrigeration. It is the residual of a calibration and not a measurement: at the base year's 1 817 kt of nitrogen and a 34% domestic share the fertiliser term is 751.6 kt, and 148.6 is what has to sit beside it to reproduce the 900 kt the retired ammoniaProduction slider asserted. That the two land within a fifth of a kilotonne of each other is a cross-check rather than a fit — the 34% comes from the fertiliser industry, the 1 817 kt from Citepa, and the 900 kt from a teaching workbook that knew neither. What would replace this constant is French ammonia output itself and its split between fertiliser and other chemistry, which the USGS does not itemise for France and which a market report's 632 kt for 2023 does not settle.

• France Fertilisants / UNIFA — 34% of French nitrogen made in France
• USGS, Mineral Commodity Summaries 2026, nitrogen — France is not itemised

34% of the nitrogen French farms use is made in France, 24% comes from other EU countries and 42% from third countries. It is the value ammoniaDomesticShare is read at where the module is switched off, and it is also that lever's default.

• France Fertilisants / UNIFA — supply shares of French nitrogen fertiliser

Cattle 38.64 + pigs 2.47 + poultry 0.23 + other livestock 4.36 in 2024 — 59% of the whole agriculture sector, and 86% of its methane. It is the calibration target of the livestock block: the three cattle factors are fitted to it in the IPCC Tier-2 order, and the other three are derived from their own lines.

• Citepa, SECTEN 2026, Agriculture-Emissions — livestock subtotal 45.70 MtCO₂e in 2024

Mineral fertilisers 9.90 + organic 1.48 + grazing 1.45 + residue burning 0.02 + other crop emissions 8.25 in 2024 — 27% of the sector, and 81% of national N₂O. The module reproduces 21.09 against it, because the published mineral line of 9.90 is its own two gases, 7.65 and 2.24, rounded up: the module adds the gases and the inventory rounds the line, and the hundredth between them is reported rather than absorbed.

• Citepa, SECTEN 2026, Agriculture-Emissions — crops subtotal 21.10 MtCO₂e in 2024

Methane yield of the wet feedstocks a digester takes — manure, crop residues and grass alike, which is how the source publishes it, as one number rather than three. Cover crops get their own, higher figure (cive_biogas_yield), because a whole green plant digests better than straw or a slurry does. It is shared rather than national because it is a property of the substrate, not of the country: the same tonne of dry manure yields the same methane in Germany. What is national is how many tonnes there are, which is manure_dm_per_cattle_head and its neighbours.

• IGEDD/CGE/CGAAER, Évaluation du potentiel de production d'énergies renouvelables à partir de la biomasse, déc. 2023, annexe 10 p. 134 — 2,0 MWh PCI/t MS pour effluents, résidus et herbe

A winter intermediate crop harvested whole gives 250–320 Nm³ of methane a tonne of dry matter, and a normal cubic metre of methane is 9.97 kWh PCI, so the range is 2.5–3.2 MWh/t DM. 2.8 is its middle, and it is the figure that makes the mission's own arithmetic work: 6 t DM/ha × 2.8 is 17 TWh per million hectares, which is what the report quotes.

• IGEDD 2023, p. 22–23 et annexe 10 — 250–320 Nm³ CH₄/t MS, 17 TWh/Mha de CIVE
• AILE, Les substrats en méthanisation — pouvoirs méthanogènes par substrat

A tonne of dry residue turned into a second-generation liquid fuel by the thermochemical route yields the same 2.0 MWh as the same tonne turned into biogas, which is exactly why the two compete: choosing one forecloses the other at no gain in energy. Declared separately from the biogas figure so that a source which does separate them can move one without the other. The equality is not a coincidence of rounding — Fischer-Tropsch converts at up to 50% and a digester's methane at a similar order — but it is a coarse number, and the route's real efficiency depends on the plant.

• IGEDD 2023, annexe 10 — 2,0 MWh/t MS en voie thermochimique ; Fischer-Tropsch ≤ 50 %, FAME ≤ 70 %

Two administrations publish two numbers and the game has to pick one. The energy directorate's biomass balance uses 2.14 MWh per cubic metre of roundwood; the environment inspectorate's annex uses 2.4. 2.14 is taken because it is the figure the balance that also supplies non_forest_wood and waste_wood is written on, and mixing two conventions inside one total would be worse than choosing the lower of them. The difference is 12%, or about fourteen terawatt-hours on the base year — larger than any single lever in this block moves. It is shared rather than national because it is a conversion, not a measurement of a country: what varies across borders is the species mix behind it, and a country that knows its own should say so in the why of the entries that use it.

• SGPE, Bouclage biomasse : enjeux et orientations, juillet 2024, p. 19 — 2,14 MWh/m³
• IGEDD 2023, annexe 7 — 2,4 MWh PCI/m³, la valeur non retenue

Back-cast from the mission's own 2050 case: 7.8 Mt DM of manure reaching digesters on a herd 30% smaller in cattle and 16% smaller in pigs. At those herds, 0.60 t DM per head of cattle and 0.08 per pig reproduce 7.71 Mt, which is the mission's figure to within a per cent. It is collectable manure — what a store receives — and not everything the animal produces: French cattle are at grass for a large part of the year and what they leave in the field never reaches a digester. The housing period is therefore already inside this coefficient rather than declared as a separate share, which is a simplification worth naming: a scenario that housed the herd differently would need a different number and this model gives it no way to say so.

• IGEDD/CGE/CGAAER 2023, annexe 10 — 6,5 Mt MS de fumier + 1,3 Mt MS de lisier en 2050, soit 16 TWh
• docs/agriculture/bioenergy.md §4.2 — le rétro-calcul et ses hypothèses de cheptel

The other half of the same back-cast. A pig is housed all year, so almost all of its slurry is collectable; what makes the figure small is that slurry is 5–13% dry matter against a solid manure's 12–23%. The whole French pig herd contributes about 0.95 Mt DM against cattle's 9.87.

• AILE, Les substrats en méthanisation — teneurs en MS des fumiers et lisiers
• docs/agriculture/bioenergy.md §4.2

Ten per cent of the collectable manure, which is 2.17 TWh of the 24.25 the base year consumed. It is not a measurement and it is half of gap 3. The mission counts about 1 Mt DM of manure going to energy in 2020, which is 9% of this model's collectable pool; the national strategy's "22% methanised by 2030" implies little more today; a gas-network panorama's site-type shares — four fifths of injected biomethane from farm sites — imply a great deal more. Both cannot be right, and no feedstock tonnage survey was obtained to settle it. It is deliberately not manure_methanised_base, which is zero. That one belongs to the emissions side: the abatement lever books the change from observed emission factors which already contain the digesters the base year ran, so booking a base-year share there would double-count the abatement. This one belongs to the supply side, where those same digesters made methane somebody consumed. Two base years, two purposes.

• IGEDD/CGE/CGAAER 2023, p. 6 — ~1 Mt MS d'effluents à l'énergie en 2020 (~1,5 TWh)
• SER/GRDF/GRTgaz/Teréga, Panorama des gaz renouvelables 2024, p. 1 et l. 710–720 — production par type de site
• SDES/Panorama, enquête de tonnages d'intrants ; observatoire de la méthanisation de l'ADEME. NON OBTENUS — c'est ce qui fermerait la lacune

Six tonnes of dry matter a hectare, INRAE's central figure. The published range is 4 to 8, and below 4 the crop is not worth harvesting; the Arvalis trial network reads 6 in the north-east and more than 10 in the south-west. The interannual spread is ±5 Mt DM on a 15 Mt national total — a third of the term, and larger than most levers in this module. Two things it assumes and this model does not charge: the crop needs some mineral nitrogen in most years, and its soil-carbon effect is positive only if the digestate goes back to the field.

• IGEDD/CGE/CGAAER 2023, p. 22 — 4 à 8 t MS/ha, 6 retenu (INRAE)
• Arvalis, réseau RECITAL, via clesdelatransition.org — 6 (NE) à >10 (SO) t MS/ha
• docs/agriculture/bioenergy.md §4.2 et §5

About 150 000 hectares in the base year, and it is an estimate rather than a census: what is published is that cover crops are 13% of the ration French digesters eat (Arvalis, 2022, rising), and 0.15 Mha at 6 t DM and 2.8 MWh/t is 2.5 TWh of the 24.25 the base year consumed, which is consistent with that share. No survey of the area itself was found.

• Arvalis via IDDRI, Le biométhane en France, oct. 2025 — CIVE 13 % de la ration des méthaniseurs (2022)
• docs/agriculture/bioenergy.md §2.3

Four million hectares, INRAE's own estimate of the French spring cropping that could carry a winter cover crop at all — the expertise reads it as 37 TWh of biogas at 6 t DM a hectare. It bounds nothing: civeArea stops at 3.0, which is what the mission recommends, so the headroom this reports is never negative inside the declared slider and is a diagnostic rather than a constraint. It is here so a reader can see how much of the physical limit the game's own maximum represents — three quarters of it.

• INRAE, expertise Biomasse-énergie 2050, citée par IGEDD 2023 note 40 — 4 Mha × 6 t MS/ha = 37 TWh
• docs/agriculture/bioenergy.md §4.2

Solagro's 57 Mt DM of French crop residues, over this model's arable area of 17.2648 Mha. The note derives 3.36 t DM/ha instead, over Agreste's 16.948 Mha of terres arables; the two differ because the land account is in Teruti hectares and Agreste's arable is a narrower class. What is measured is the tonnage, so the tonnage is what is preserved: at the base year this coefficient reproduces 57.0 Mt DM exactly, where 3.36 would have produced 58.0 and quietly added a megatonne of straw nobody counted.

• Solagro, Quelles biomasses pour la transition énergétique ?, oct. 2024, p. 20 — 57 Mt MS de résidus
• docs/agriculture/bioenergy.md §4.2 — la valeur 3,36 t MS/ha et son dénominateur

One per cent of the residue pool leaves the field for energy today — about 0.57 Mt DM, or 1.1 TWh split between biogas and second-generation liquid. It is an estimate: what is published is that straw export is marginal against the tonnage produced, not a national figure for it. The lever's default of 16% is therefore a sixteen-fold increase, which is worth knowing before reading the residue term as an easy gain.

• INRA/ADEME (AGRICE), Prélèvement de paille — l'export actuel est marginal devant le surplus
• docs/agriculture/bioenergy.md §4.2

Half the mobilised residue goes to a digester and half to a second-generation liquid plant. The mission's own 2050 case splits 5 Mt DM to biogas against 10 Mt DM to liquid, which is a third; half is taken here because the mission's split is a recommendation about which industry to build rather than a property of the straw, and because a fixed half makes the competition between the two pools legible on the dashboard — raise the mobilisation lever and both bands move together, which is the teaching. What matters more than the value is that the split is exhaustive: this share and its complement are the only two claims on the pool, so the same tonne cannot be counted in both, and a test asserts it.

• IGEDD/CGE/CGAAER 2023, annexe 10 — 5 Mt MS au biogaz, 10 Mt MS au 2G
• docs/agriculture/bioenergy.md §5 — « choisir l'une ferme l'autre »

Nineteen terawatt-hours, 78% of the base-year biogas total, and the module's largest declared hole. It is what closes 2024 once the manure, the cover crops and the residues this model builds are counted: household biowaste, sewage sludge, agri-food effluent and landfill gas, plus whatever share of the feedstock the published sources disagree about. The disagreement is the point. The national strategy's "22% of manure methanised by 2030" implies very little manure in today's digesters; a gas network panorama's site shares — four fifths of injected biomethane from farm sites — imply a great deal. Both are published, both cannot be right, and no feedstock tonnage survey was obtained. So the residual is declared, it moves with no lever, and every build prints it with its share so it cannot quietly become the answer. An SDES or Panorama feedstock survey, or ADEME's methanisation observatory, would close it. Read the biogas supply with this number in mind: at the reference the module produces about 70 TWh, and 19 of them are this.

• SDES, Chiffres clés des énergies renouvelables 2025, feuille P3_G25 — 24,25 TWh de biogaz en 2024, tous usages
• SER/GRDF/GRTgaz/Teréga, Panorama des gaz renouvelables 2024 — parts par type de site
• SDES/Panorama, enquête de tonnages d'intrants ; ADEME, observatoire de la méthanisation. NON OBTENUS — la lacune 3 de la spécification du module

Of every cubic metre that leaves the French forest as material, 58% comes back as fuel: sawmill offcuts and bark, panel residues, and the black liquor a pulp mill burns to run itself. It is fitted so the base-year wood supply lands on the 120.05 TWh the energy statistician observes, and it is the second-largest thing in that total — about 35 TWh. It is large because sawing is wasteful: a log yields well under half its volume as sawn timber, and the rest is chips, slabs and sawdust that a boiler or a panel press takes. A sawnwood balance against sawlog input, with the paper industry's own energy consumption beside it, would measure it instead of fitting it.

• SDES, Chiffres clés des énergies renouvelables 2025 — 120,05 TWh de bois-énergie en 2024, la cible du calage
• Agreste, Récolte de bois et production de sciages en 2024 (Chd2517) — le bilan qui la mesurerait
• docs/agriculture/bioenergy.md §2.2 et §4.3

Hedges, orchards, vineyards and trees outside woodland, burned for energy. 22.8 TWh is the national energy and climate plan's figure as the inspectorate reports it; ADEME's 2017 count is 25.7. The lower is taken, for the same reason wood_energy_per_m3 takes the lower of its two figures: it is the number the balance the rest of this block is written on uses. It is a fixed term because no lever in this model plants or grubs up a hedge — a real limitation, since hedgerow policy is one of the few biomass decisions a farm can take on its own.

• IGEDD/CGE/CGAAER 2023, annexe 7 — 22,8 TWh (SFEC) ; 25,7 TWh (ADEME 2017)
• docs/agriculture/bioenergy.md §2.2

About 2 Mt of dry matter of end-of-life wood — pallets, packaging, demolition timber — burned for energy, which is 9 TWh. An older energy directorate count puts it at 16.5 TWh on a wider perimeter; 9 is the inspectorate's own figure and is on the perimeter of the rest of this block. A fixed term: it is a waste-management quantity, and this model has no waste-management lever that reaches it.

• IGEDD/CGE/CGAAER 2023, annexe 10 — 2 Mt MS de bois déchet ≈ 9 TWh (16,5 TWh dans le comptage DGEC 2019)
• docs/agriculture/bioenergy.md §2.2

The average of the crops France actually grows for fuel, weighted by their areas: wheat 115.8 kha and maize 75.2 kha at 18 MWh/ha of ethanol, sugar beet 27.2 kha at 53, and about 400 kha of rapeseed and sunflower at 17 of FAME. That is 11.68 TWh on 618 kha, or 18.899 MWh a hectare. Two things it makes visible. A beet hectare is worth three oilseed hectares, so the mix matters as much as the area — and this model holds the mix fixed while the area moves, which is a simplification a scenario that wanted more beet would have to break. And 11.68 TWh reproduces the mission's "about 11 TWh from French land" by a route that starts from hectares rather than from fuel, which is a cross-check rather than a fit. The weakest input is the 400 kha of oilseed: a FranceAgriMer estimate quoted by the trade press, not verified at source, carrying 6.8 of the 11.68 TWh.

• FranceAgriMer, Fiche filière bioéthanol 2026, p. 1 — blé 115,8, maïs 75,2, betterave 27,2 kha
• FranceAgriMer, Veille concurrentielle biocarburants 2023 (éd. 2024), p. 13 — 1,5 à 2 t d'huile/ha de colza, soit 15 à 20 MWh/ha
• IGEDD/CGE/CGAAER 2023, p. 6 et annexe 10 — ~11 TWh de 1G attribuables à la terre française
• docs/agriculture/bioenergy.md §2.4 et §4.2 — le détail par culture et l'avertissement sur les 400 kha

218 kha of ethanol crops, published by crop, plus about 400 kha of oilseed, estimated. 0.8% of the utilised agricultural area, for 11.7 TWh — against 41.7 TWh of liquid biofuel the country actually burned. The gap is imported, as fuel or as feedstock, and that is the number this constant exists to make readable.

• FranceAgriMer, Fiche filière bioéthanol 2026 — 218 kha, 0,8 % de la SAU
• docs/agriculture/bioenergy.md §2.4

Used cooking oil is 5.3% of French biodiesel and animal fats 4.7%, which is 3.0 TWh of the domestic supply. Solagro's ceiling is 0.4–0.5 Mt DM, or 5–6 TWh: the quantity is set by how much a country eats and slaughters and no lever can raise it, which is why it is a constant and not a slider. It is also the one liquid route whose life-cycle emissions are genuinely low — 54 gCO₂/kWh against 179 for rapeseed FAME under the renewable energy directive's defaults — and the model does not distinguish it, charging every liquid at efLiquid.

• SDES, Chiffres clés des énergies renouvelables 2025, feuille P3_G30 — HAU 5,3 %, graisses animales 4,7 % du biodiesel
• Solagro, Quelles biomasses pour la transition énergétique ?, oct. 2024, tableau 2 — 0,4 à 0,5 Mt MS, plafond dur

What France imported in the base year, as finished fuel and as feedstock, derived so the supply check closes on the observed total: 41.7 TWh consumed less the 15.25 TWh this model builds from French land, waste fats included. The planning secretariat's own 2023 balance reads 19 TWh of fuel plus 11 TWh of feedstock imported against 14 exported, which is the same order by an independent route; the international energy agency reads net imports at 48% of biodiesel use and 30% of ethanol. It is the base year of bioImports and the only anchor its bounds have.

• SGPE, Bouclage biomasse, juillet 2024, p. 18 — 34 TWh de matière première (11 importés) → 35 produits (14 exportés) → 40 utilisés (19 importés)
• IEA Bioenergy, Implementation of bioenergy in France, 2024 — importations nettes biodiesel 48 %, éthanol 30 %

Primary consumption of wood energy in 2024, 29.5% of all French renewable energy and the largest single renewable the country has. It is the target wood_byproduct_share is fitted to, and the check the tests run. Not climate-corrected, and that matters for a term two thirds of which is household heating: a mild winter moves this number by several terawatt-hours without anything physical changing.

• SDES, Chiffres clés des énergies renouvelables 2025, feuille Données_clés_G1 — bois-énergie 120,05 TWh, 29,5 % des EnR

Primary consumption of biogas in 2024, all uses on the lower heating value: 10.44 TWh injected into the network (11.6 TWh on the higher value the gas industry publishes), 7.99 in cogeneration, 1.98 in electricity alone and 0.03 in heat alone. It is the target biogas_other closes on, so the share that residual represents of this number — 78% — is the module's headline gap.

• SDES, Chiffres clés des énergies renouvelables 2025, feuille P3_G25 — biogaz 24,25 TWh en 2024, par usage

Primary consumption of liquid biofuel in 2024 — biodiesel 73%, bio-gasoline 25%, bio-jet 2% — imports included. The same publication's filière table reads 41.05 on a slightly different perimeter, and the gross final consumption of transport is 38.5; 41.7 is taken because it is the primary figure, which is the basis the other two pools are on. It is the only one of the three base-year checks that is not closed by a calibrated residual, so it is the one that says whether the liquid coefficients are right. It lands at −0.05 TWh, which is a tenth of a per cent.

• SDES, Chiffres clés des énergies renouvelables, édition numérique, page « Biocarburants » — 41,7 TWh en 2024, biodiesel 73 %, bio-essence 25 %, bio-jet 2 %
• SDES, Chiffres clés des énergies renouvelables 2025, feuille P3_G29 — 41,05 TWh sur le périmètre filière